CHARACTERS USEFUL AT HIGHER LEVELS OF RELATIONSHIP



*Click Here* and Character List will appear to your left.

Below I discuss briefly major characters in the order in which they will be encountered in the node and family characterisations. The emphasis here is on providing entries both to the general morphological literature, but especially to articles where there is much information on variation of a single character; I also give indication of how good the sampling is. Definitions of the characters mentioned here and of related features are provided in the Glossary. Note, however, that these characters and the "states" I use in describing the variation within them should not simply be converted into data for phylogenetic analysis. Some, for example, fruit types, are logically connected with the nature of the gynoecium, others, such as seedling morphology and testa and tegmen types, might each yield a dozen characters, while in these and others (all too many) the states are arbitrarily delimited. Finally, it should be remembered that because the same term is used for similar features of two plants, it cannot be assumed that this entails a hypothesis that the structures in the two organisms had a common evolutionary origin, i. e., that they are homologous under one definition of the term. However, if botanical terms were reformulated so as to refer only to synapomorphies (shared derived features), changes in the terms we use would be almost innumerable!

There are of course many other features mentioned in individual family characterisations. For these, general distributions may be unknown or not really of much interest because the feature is varying in a phylogenetically informative fashion only locally, or the feature may be unique or almost so.

GENERAL

I have necessarily relied much on information in Airy Shaw (1966), Cronquist (1981), Engler (1900-onwards) and Engler and Prantl (1887-1915; ed. 2, 1924-onwards - cited under individual authors), Goldberg (1986, see also 2003), Hutchinson (1973), Mabberley (1997), Takhtajan (1997: beware of typographical errors!), and N. Smith et al. (2004). The meticulously-documented series of papers that make up the Generic Flora of the Southeastern United States (ed. C. E. Wood Jr., e.g. Wood 1983) and in particular the series of volumes edited by Kubitzki (1993a onwards) have been invaluable; most papers are cited separately in the individual family accounts. Much information is summarised in Rolf Dahlgren's diagrams (e.g. Dahlgren 1975) and is included in recent large-scale phylogenies (e.g. Hufford 1992: Rosidae; Nandi et al. 1998: all flowering plants). Zomlefer (1994) was helpful, especially early on. Information on many characters mentioned below is taken from these references and also from the general references mentioned at the beginning of each character group; this is definitely not the best method of documentation (see Introduction).

Habit/life form (see also inflorescence position). This is indicated in a very general fashion; for an interesting survey of plant life forms, see Raunkiaer (1934). Groover (2005) discusses the multiple derivations of the herbaceous life form from the woody, and the reverse; the latter can often be detected by details of xylem anatomy, and is discussed under individual families. The growth of monocots, predominantly a perennial herbaceous group, can be interpreted as variation on a basically sympodial theme (Holttum 1955). Procheş et al. (2006) discuss the richness of the geophytic flora in the Cape region of South Africa, at the same time observing how hard it is to define what a geophyte was...

Leake (1994) surveyed the biology of echlorophyllous myco-heterotrophic plants, observing that they were commonest in monocots. Note that mycotrophic plants often lack stomata and are otherwise distinctive anatomically; they often also have parietal placentation, many ovules with long funicles, small seeds, little endosperm, an undifferentiated embryo, etc.

The only angiosperms that directly parasitize other angiosperms are broad-leaved angiosperms (Leake 1994), such parasitism having evolved ca 12 times (Barkman et al. 2007: the mitochondrial coxI gene has acquired an intron in ten of these cases; Westwood et al. 2010); there is a single gymnosperm, the podocarp Parasititaxus usta, that is parasitic - on another podocarp (Qu et al. 2019 and references). Like myco-heterotrophic plants, the floral morphology of parasitic plants is often very distinctive and difficult to relate to that of their putative photosynthetic relatives (see Heide-Jørgensen 2013 and references for parasitic syndromes). They have a distinctive physiology, too; for instance, being a parasite often seems to entail having a high rate of transpiration both day and night (Stewart & Press 1990).

Plant architecture. Hallé et al. (1978) list a number of models, growth patterns based on variation in a number of characters and to which plants may be assigned (see also Hallé 2004). These are rarely mentioned here except in passing since they are not often constant in taxa of any size. However, some of the individual characters that Hallé et al. (1978) discuss, e.g., ortho- and plagiotropy (the latter in particular comes in various flavours), continuous and discontinuous (rythmic) and monopodial and sympodial growth, are mentioned below. I find thinking of plant growth in terms of these variables very helpful, much more so than thinking of models. Branching is generally axillary, but a number of aquatic Alismatales show vegetative bifurcation of the vegetative axes, perhaps precocious axillary branching; strictly dichotomous branching is very rare in seed plants, occuringg only if Hyphaene (Arecaceae), Strelitzia (Strelitziaceae), and a few other examples (e.g. Wilder 1975). Leaf insertion (also see below) varies independently of plant architecture.

Dessication tolerance. Although this is basically a physiological character, a relatively few plants, usually more or less herbaceous, show various degrees of dessication tolerance, some being known as "resurrection plants", examples being Myrothamnaceae, many Velloziaceae (Alpert & Oliver 2002; Proctor & Pence 2002; Dickie & Pritchard 2002).

Nitrogen fixation. Angiosperms that can fix nitrogen in association with bacteria are uncommon, yet are found almost exclusively in a group of four orders, the N-fixing clade (Fabales, Rosales, Cucurbitales, Fagales). Jeong et al. (1999) and Clawson et al. (2004) discuss this association in the context of the evolution of Frankia, the bacterium most commonly involved outside the Fabaceae.

CHEMISTRY

Hegnauer (1962 onwards) is the source of information, and Siegler (1998) is also invaluable. Gershenzon and Mabry (1983) can still be read with profit, Wink (1999) discusses the biochemistry of plant secondary metabolism that yields the probably in excess of 100,000 secondary metabolites found in flowering plants, the distribution of a number of which are of systematic interest, Kite et al. (2000) provide a useful survey of the distribution of some chemicals in the monocots, while Strauss and Zangerl (2002) summarise the diversity of the major classes of compounds involved in chemical defence in angiosperms (see also insect-plant and fungus-plant relationships mentioned at the end of this page where secondary metabolites are commonly involved). Waterman (2005) provides a general survey of the diversity of secondary metabolites. Note that sampling for chemical characters may be spotty (although often better than that for many embryological characters?), and it also can be difficult to say that a compound is really absent. Thus myricetin occurs in Magnoliaceae and Annonaceae (Crawford et al. 1986), although its apparent absence there had earlier been a matter for comment (Kubitzki & Reznik 1966; Gornall et al. 1979). Almost all authors rightly emphasize the importance of the pathways by which compounds are produced; apparently distinctive substances like napthoquinones can have quite diverse biosynthetic origins, yet at the same time furanocoumarins in Ficus, Apiaceae and Rutaceae - quite unrelated - are produced by the same pathway (Berenbaum 1983 and references). In general, chemical characters are like others in terms of their value in phylogenetic construction; some, like the presence of glucosinolates, are very valuable, many others, less so. Still others - presence of ellagic acid is an example - show a fairly resticted distribution, but are still very variable in terms of presence/absence (D. Soltis et al. 2005b). Monocots in general have a less diverse secondary chemistry than do other angiosperms.

Alkaloids. For a general introduction, see Hegnauer (1988). For a survey of the structure and distribution of the isoquinoline alkaloids, see Buck (1987) and Bentley (1998), while Liscombe et al. (2005) put the origin of these alkaloids in a phylogenetic context. Alkaloids are extremely diverse (some 10,000 different kinds are known), and similar alkaloids can be formed via different biosynthetic pathways. Detection of the activity of (S)norcolaurine synthase, the gateway to benzylisoquinoline synthesis, may suggest that these alkaloids will be found in groups in which they are not currently known, e.g. Chloranthaceae. For the synthesis of benzylisoquinoline alkaloids in the opium poppy (they are synthesised in the sieve tubes!), see Bird et al. (2003); these are normally, but not always, derived from tyrosine. For pyrrolizidine alkaloids, see Ober and Hartmann (2000) and Anke et al. (2004), also Boraginaceae, Apocynaceae and Asteraceae. The distribution of some distinctive types such as the erythrina (Erythrina, Cocculus) and homoerythrina (Phelline, Dysoxylum, Schelhammera, some conifers) alkaloids is very scattered. For the distribution of calystegines (tropane alkaloids) see Dräger (2004).

Aluminium accumulation. Useful surveys are found in Chenery (1948), Webb (1954), Kukachka and Miller (1980), and Jansen et al. (2002b). Broadley et al. (2001) summarizes patterns of heavy metal accumulation in general in angiosperms.

For calcium oxalate crystals, see below under anatomy.

Some information on the distribution of chelidonic acid is taken from Ramstad (1953).

Coumarins are involved in plant defense, some 800 different kinds being known form about 80 families (see also Berenbaum 1983).

Cyanogenetic pathways - see the survey by Saupe (1981) and comments by Hegnauer (1986). Some 60 different kinds are known from about 130 families.

Ethereal oils are more or less volatile and aromatic; they are often terpenes of some sort. They may be found in idioblasts, i.e. specialised cells, as in basal angiosperms, and leaf blades, tepals, etc., containing ethereal oil cells may be punctate when a light is shone through them.

Flavonoids. For general information about flavonoids, see Giannasi (1988), Bohm (1988), and Harborne and Baxter (1999), for minor flavonoids, see Bohm (1988), for isoflavonoids in particular (often phytoalexins), see Reynaud et al. (2005) and Mackova et al. (2006), for biflavonoids, Geiger and Quinn (1975), and for flavonoid distribution in monocots, Williams and Harborne (1988). As Stafford (1990: p. 9) notes, "the use of duplicate or triplicate names for the same group of flavonoids is an unfortunate aspect of flavonoid nomenclature". Hegnauer (1986) suggests that flavonoids are of little use when looking at higher-level relationships. Anthocyanins are common pigments, and when found in leaves they are usually red, perhaps protecting the young leaf (Manetas 2006: insect herbivores cannot see red, and if they are green, they contrast with leaf color).

Fructose oligosaccharides are sometimes stored by the plant; families in which they have been detected are listed by Pollard and Amuti (1981). Such oligosaccharides are mentioned only when they occur reasonably consistently within a family - so they are not mentioned in Cornaceae, where they occur in a few species of Cornus, but not others, nor in Melianthaceae, where they are reported only in trace amounts. For the distribution of stem fructans in monocots, see Pollard (1982).

Glucosinolates are largely restricted to Brassicales (Fahey et al. 2001 for a summary). Some 100 different forms are known. They are broken down by the enzyme thioglucoside glucohydrolase (myrosinase) into glucose and isothiocyanates (mustard oils - R-N=C=S arrangement). Mustard oils are uncommon in flowering plants (see also Putranjivaceae - Malpighiales), and their presence is systematically important. For details of glucosinolates and their distribution, see Fahey et al. (2001: note that some of the more unexpected records of glucosinolates, e.g. in Phytolaccaceae and Pittosporaceae, are reported by Daxenbichler et al. 1981, however, there are no vouchers and the reports must be considered suspect - J. E. Rodman pers. comm.), for their ecological and agricultural implications, see Courtney (1986), Chew (1988), Brown and Morra (1997) and Rask et al. (2000), and for herbivores that are attracted to mustard oils, see Host-plant preferences below.

Gums, non-crystalline mixtures of sugars and organic acids, are scattered in angiosperms, being notably common in Fabaceae, where they are often found in the seeds (Nair 1995).

Iridoids. Data are taken mainly from Bate-Smith et al. (1975), Hegnauer and Kooiman (1978), and especially Jensen (1991, 1992, 1999) and Jensen et al. (1975b). Iridoids are formed from two isopentane (isoprene) units, and they are cyclic C10 compounds, often (always?) with O in one ring; they are formed in two main ways; they may be route I (secoridoids) or route II (carboxylated and decarboxylated iridoids: see Glossary). Although a very useful character, presence/absence of iridoids may vary in closely related taxa (e.g. Jensen et al. 1975b). (Note that Inouye and Uesato [1986] question classifying iridoids based on their biosynthetic pathways.) Iridoids are produced by the mevalonate pathway, and although polyacetylenes have a very different origin, being derived from their corresponding fatty acids, in familes or groups of families of asterids one or other of these two classes of compounds dominate to the exclusion of the other. Families that contain iridoids are largely asterids, although there are a few in Saxifragales, etc.; all told, some 600 different structures are known. For further details of possible evolutionary pathways of iridoids, see Albach et al. (2001a). As glycosides, many have distinctive, often adverse, physiological effects on mammals, micro-organisms, etc. Bowers (1988) summarises aspects of the relationships between iridoids, plants and insects and Nishida (2002) examples of insects that sequester the compound; see also Host-plant preferences below.

Iron uptake. The roots of most plants extrude protons which lower the pH of the adjacent soil and solubilize iron. Those of grasses, however, produce siderophores which chelate ferric ions and are then subsequently taken up by the plant (Schmidt 2003). I do not know the exact distribution of this character.

Latex is mentioned only when it occurs (see also laticifers, below). Gutta has the trans- configuration of isoprene units, and is distinguished from latex, which nas the cis- configuration, in the descriptions.

Lignans and neolignans. Lignans are ß,ß'(8-8')-linked dimers of cinnamic acid residues or their biogenetic equivalents (diaryl propanoids), and are widespread, although they apparently do not occur in monocots. They are also found in gymnosperms. Neolignans, more directly linked, or at least other than ß,ß', are found especially in magnoliids (monomeric allyl- and propenylphenols are also found in the groups where they occur). For definitions (which are not consistent, and depend either on structure or on biosynthetic pathway), distribution, etc., see McRae and Towers (1984), Whiting (1985: structure) and Ayres and Loike (1990).

Lignins are complex and still poorly understood components of secondary cell walls. Their composition is of considerable interest, and under certain conditions they yield substantial amounts of aromatic aldehydes - various combinations of vanillin, syringaldehyde, and p-hydroxybenzaldehyde. In particular, the Mäule reaction, in which a reddish colour is obtained after treatment of the plant sample with potassium permanganate, then dilute hydrochloric acid, and then ammonia, signifies presence of syringaldehyde (Gibbs 1957 for an early but fairly comprehensive survey), with the ratio of syringyl to guaiacyl units being less than 2-2.5:1. The occurence (or otherwise) of this reaction is correlated with fairly major taxonomic groups; note that vascular tissue may not always be stained by this reaction (and even when it is, it does not always stain primary xylem). Vanillin is a much more widespread component of lignin, but p-hydroxybenzaldehyde is also somewhat restricted, being absent from broad-leaved angiosperms - at least from magnoliids and eudicots (Towers & Gibbs 1953). For a general discussion on details of the synthesis of lignin precursors, see Li et al. (2001), and for discussion on the distribution of lignin types, see Harris (2005).

Mannitol. This is particularly common in groups like Oleaceae, but the systematic significance of its distribution is unclear (see Stoop et al. 1996).

The metabolic pathway of nicotinic acid in tissue culture may be of systematic significance, although at some 50 species the sampling is poor (see Willeke et al. 1979). Gymnosperms and all flowering plants except asterids metabolise nicotinic acid into trigonelline.

Non-protein amino acids are mentioned only when they are abundant; they are quite widely distributed, and some 600 different forms are known.

Phenols are widespread. The distribution of caffeic acid, perhaps involved in plant defence, seems to be of some systematic interest. It is especially common in herbaceous groups, and is produced from cinnamic acid; proanthocyanadins or non-hydrolysable tannins (see below under tannin) are also produced from cinnamic acid and tend to show the inverse distribution.

Phorbol esters are diterpenes, often with a toxic effect (Evans & Taylor 1983 for distribution; Evans & Edwards 1987 for physiological effects [for which also see other papers in volume 94 of the Botanical Journal of the Linnean Society]).

Photosynthetic pathways. There are three main pathways by which carbon is incorporated into the organism or, perhaps more accurately, is initially captured prior to being incorporated into metabolic cycles, the C3 and C4 pathways, also CAM (Crassulacean Acid Metabolism). Recent work (Surridge 2002; Sage 2004 for references) emphasises that C3 and C4 pathways may occur in the same organism, and the spatial separation of the different parts of C4 photosynthesis may be at the scale of the cytoplasm and chloroplast within a single cell, as well as in different cells organised as distinct tissues (see also Bowes et al. 2002; Voznesenskaya et al. 2003 - more examples of separation at the subcellular scale are being reported). For a survey of the C4 pathway, see Sage and Monson (1999). Despite the recent discovery of dinosaur coprolites from the Late Cretaceous (71-65 mybp) of central India (Prasad et al. 2005) with contained fossils of derived grass clades, C4 photosynthesis in Poaceae seems to have originated in the middle Miocene, some 12.5 mybp, and it expanded greatly as recently as 9-4 mybp (Jacobs et al. 1999: palaeoecology of Poaceae; Osborne & Beerling 2006: the general evolution of C4 photosynthesis). For literature on CAM, see Lüttge (2004, 2005). Although distinctive pathways are rarely constant in the taxa characterised below and have evolved in parallel many times, they are particularly common in groups like core Caryophyllales, Poales in general and Poaceae in particular, etc.

It might seem that phytochrome behaviour is perhaps of interest phylogenetically; very limited evidence suggests that Pr is converted to Pfr, and then in core Caryophyllales, Pfr undergoes light-dependent decay, while in broad-leaved angiosperms it reverts to Pr in a thermal dark reaction (Kendrick & Hillman 1971). However, Sarah Mathews (pers. comm.) observes that light-dependent decay of Pfr is very widespread indeed in angiosperms.

Polyacetylenes are derived from their corresponding fatty acids, but in asterids either polyacetlyes or iridoids dominate to the exclusion of the other even although they are biosynthetically unrelated.

Quaternary Ammonium Compounds such as glycine betaine commonly occur in plants growing in saline habitats, and are probably involved in maintaining the osmotic balance of cells and organelles (Rhodes & Hanson 1993). Tertiary sulfonium compounds may also be involved. However, details of the distribution of glycine betaine and related compounds are still not well understood, and they are also common in many Lamiaceae (Blunden et al. 1996) and Convolvulaceae - families not noted as being particularly halophytic.

Resins and substances either similar to or confused with them are quite widespread in plants; for their distribution, see see Nair (1995), and for a readable summary, Langenheim (2002).

For suberin, see Bernards (2002).

Tannins are either condensed and non-hydrolysable (built up of proanthocyanidins, for which see Dixon et al. 2005) or hydrolysable (ellagic acid, also gallic and caffeic acids, are the building blocks). The two kinds of tannins are not chemically immediately related, and although the unspecific terms "tannins" and "tanniniferous" occur in some of the characterisations, they are close to useless; the generic terms non-hydrolysable and hydrolysable tannins are to be prefered. proanthocyanidins and caffeic acid tend to be alternativesand the latter is especially common in herbaceous taxa, e.g. the asterid I and II groups (Mølgaard 1985 and Mølgaard & Ravn 1986, which see for more details than have been incrporated here). Early surveys by Bate Smith (1962) and Bate Smith and Metcalfe (1957) are still useful.

Terpenoids are a diverse and ecologically important group of compounds (e.g. Harborne & Tomas-Barberan 1991) made up of isoprene building blocks (seee Bohlmann et al. 1998 for early steps in synthesis). They are found in all plants, and some 15,000 different kinds are known.

Terpenoids are related to steroids, etc.

VEGETATIVE ANATOMY

Metcalfe and Chalk (1950; 2nd edition, 1979 onwards, incomplete) provide information on general anatomy, although characters like nodal anatomy are rarely mentioned in the first edition; classical literature remains useful (see e.g. Thouvenin 1890; Beauvisage 1920; Watari 1939 for taxa that used to be placed in Saxifragaceae). Gregory (1994) is an invaluable bibliography with the more important contributions conveniently indicated, Dickison (2000) is a very accessible introduction to general plant anatomy, and Carlquist (2001) has recently updated his survey of comparative plant anatomy (see also Carlquist 1988b); Esau (1977) is still very useful. Herendeen et al (1999b) surveyed wood anatomical features in basal angiosperms and compared their findings to the classical evolutionary trends (see below). Wheeler et al. (1989) provide a useful survey of terms used in wood anatomy, while InsideWood: An Internet Accessible Wood Anatomy Database http://www.lib.ncsu.edu/insidewoood/ is a developing resource for information about wood anatomy, also identification, etc.

Carlquist (e.g. 1988, 1998b, 1998c) has emphasised the linkage between features of wood anatomy and the environment. For this reason, and also because many characters have states whose limits need justification, especially in the context of angiosperm-wide surveys such as this, wood anatomy must be interpreted with caution (for wood anatomy and cladistic characters, see Herendeen & Miller 2000). However, more information on wood anatomy is gradually being added to the characterizations.

Mycorrhizae. For the evolution and ecological significance of mycorrhizae, see Malloch et al. (1980), Read et al. (2000), etc., and for an account with a summary of the distributions of the main mycorrhizal types within angiosperms, see Brundrett (2002). Brundrett (2004) offers a classification of mycorrhizal types based on anatomical criteria regulated by the host plant. Roots with mycorrhizae commonly lack or have only a few root hairs. Ectomycorrhizae form a Hartig net of hyphae investing the rootlets and penetrating between the cortical cells; the hyphae are septate and are not intracellular - with the exceptions of Ericaceae and Orchidaceae. Bruns and Shefferson (2004) provide a convenient summary of taxa with ectomycorrhizae, but more detail will often be found in the pages here. Although basidiomycetes are frequent ectomycorrhizal associates, Pezizales (ascomycetes) are also quite common (Tedersoo 2006, ascomycetes with a hypogeous life style are derived from them). Ectomycorrhizae are commonly found in nutrient-poor and organic material-rich soils, especially in temperate and tropical montane habitats. Interesting, ectomycorrhizal associations are found in plants like Graffenrieda emarginata (Melastomataceae) in tropical montane environments along with the next kind of mycorrhiza (Haug et al. 2004 and references). Vesicular-arbuscular mycorrhizae formed by an association of Glomeromycota (Schüßler et al. 2001) with land plants - indeed, they are found only in association with plants, whether sporophyte or gametophyte. There is evidence of such mycorrhizae in the fossil record from the Silurian/early Devonian ca 400 mya, but this association has almost certainly arisen a number of times (e.g. Duckett et al. 2006). The association of Glomeromycota with flowering plants is probably of very long standing (see also Baylis 1975; Redecker et al. 2000b; etc.). Glomus itself is probably paraphyletic (or worse: Redecker et al. 2000a; Schwarzott et al. 2001) and Glomeromycota seem to be a very ancient asexual group (Pawlowska & Taylor 2004) whose basic genetic structure is not understood (Rodriguez et al. 2004). In vesicular-arbuscular mycorrhizae the aseptate hyphae are intracellular, often forming vesicles or branching structures called arbuscules within the cells (when the latter alone are formed, an arbuscular mycorrhiza may result). The various mycorrhizal associations of angiosperms are summarized by Malloch et al. (1980), F. A. Smith and Smith (1997), S. E. Smith and Read (1997), and Landis et al. (2002); for phylogenetic aspects of the associations, see also Trappe (1987); for details of the mycorrhizae in Ericaceae, see that family.

Note that a wide variety of fungi form ectomycorrhizal associations, but forests where such associations are common are not very diverse; the reverse set of relationships hold in forests with endomycorrhizal associations (Malloch et al. 1980). Aquatic plants, hardly surprisingly, lack mycorrhizae, but the frequent absence of mycorrhizae in Caryophyllales, Proteales, etc., is systematically interesting.

The issue of nitrogen uptake by the plant may be mentioned here. In ectomycorrhizae such as those associated with Eriaceae nitrogen in amino acids released by the fungus is taken up by the plant. In, say, the tundra habitat there are a number of plants that take up substantial amounts of amino acids directly from the soil, but the ability to do this is not obviously correlated either with mycorrhizal status or taxonomy. Thus in such habitats a number of Cyperaceae, which lack mycorrhizae, take up N predominantly in an organic form, but others take it up in an inorganic form (Raab et al. 1999). Arabidopsis thaliana, which also lacks mycorrhizae, also takes up N in an organic form (Hirner et al. 2006).

Endophytic fungi are found in a number of flowering plants, perhaps particularly in Poaceae (q.v.) and Ericaeae (e.g. Petrini 1986 and other references in this volume; Saikkonen et al. 2004 and references), although they are probably much more widespread. Indeed, Arnold et al. (2001) found 418 morphospecies of endophytes in only 83 leaves of two species of tropical trees (Ouratea, Heisteria). The hyphae of such endophytic fungi tend to pervade the tissues of the plant host. These associations are probably at least intermittently mutualistic (Carroll 1988, 1995).

For rusts and their hosts, see OTHER at the end of this page.

Root apical meristem. Details of the organisation of the root apical meristem are known for relatively few plants, but whether or not the tiers of initials are closed, i.e. are clonally distinct, open, with the fate of the derivatives of the apical cells not being immediately obvious, or some more or less intermediate condition intermediate-open, may be of systematic significance (Groot et al. 2004). The character has been placed on the tree, as much to draw attention to it as anything else; monocots are notably variable.

Root stele. The stele or vascular tissue in the root is central, and the xylem forms a solid mass, although it is sometimes medullated, perhaps especially in monocots. However, the root stele of a number of broad leaved angiosperms, e.g. Mimusops (Sapotaceae) and the primary roots of Clusiaceae and Dipterocarpaceae is medullated, while that of some monocots, e.g. sapromycotrophic taxa, some Alismatales, Sisyrynchium, etc. is not medullated (van Tieghem & Douliot 1888; von Guttenberg 1968). The proportion of the root occupied by the stele varies considerably, and in a small study it was found to occupy 20-48% of the area in carnivorous plants (and was sometimes medullated), ususally rather less than 34% in other plants (Adamec et al. 2006). Tracheids, and sometimes also phloem, are to be found scattered in the pith in some monocots (von Guttenberg 1968). In broad-leaved angiosperms there are commonly two to five xylem poles (di- to pentarch) alternating with phloem poles; the lateral roots arise on the xylem radii except when the roots are diarch. Diarch and tetrarch, etc., taxa commonly occur in the same family, but some larger groups (e.g. Rosales, possibly Brassicales) are overwhelmingly diarch. In diarch taxa the lateral roots arise on either side of the xylem poles, but in Pittosporaceae and their relatives origination in this distinctive position is found in taxa that are other than diarch (van Tieghem & Douliot 1888; von Guttenberg 1968). Monocots are more often polyarch (even the primary roots are not diarch), and in a number - but not all - of Poales the lateral roots arise opposite the phloem poles (van Tieghem & Douliot 1888).

Root hair development normally occurs in epidermal cells that are rather like cells that do not have hairs. However, root hair development in monocots differs substantially from that of most other flowering plants. Here there are trichoblasts, small, densely staining cells that give rise to root hairs, and these are clearly differentiated from other epidermal cells. The trichoblasts occur in vertical files, and alternating cells, usually the proximal cells (i.e. nearer the root apex), produce the root hairs, although there is infrafamilial variation in this (e.g. Poaceae, Hydrocharitaceae). In some monocots, on the other hand, especially those with a velamen, it is the hypodermal (not epidermal) cells that may be dimorphic (Kauff et al. 2000). The basic condition for angiosperms seems to be the absence of trichoblasts. However, in Arabidopsis and relatives hairs develop at the proximal ends of trichoblast cells (Schiefelbein et al. 1997), with trichoblasts arising from epidermal cells that are situated above the radial walls of cortical cells; these trichoblasts are in vertical files (Pemberton et al. 2001; Dolan & Costa 2001). There seems to be a relationship between the development of cutinized root hairs from outer cortical cells and the absence of secondary growth in the root (Pinkerton 1936: there is usually no secondary growth in monocot roots).

Root hair cellulose. Cellulose fibrils in the outer epidermal wall of the root elongation zone are normally laid down transverse to the root axis, but in some Poales they are parallel to the long axis (Kerstens & Verbelen 2002). However, Pothomorphe (Piperaceae) behaved differently, and sampling particularly in clades other than eudicots and monocots needs to be improved.

Root epidermis development. Whether or not the epidermis of the root develops from the inner layer of the root cap or the outer layer of the cortex varies at about the same level as root hair development (Clowes 2000). The inner epidermis of the roots is lost completely in monocots (and Nymphaeaceae!), the epidermis then developing from the cortex, although the inner layer of the epidermis remains attached in broad-leaved angiosperms (van Tieghem & Douliot 1888). Monocots in particular may have distinctively-thickened cells in the exodermis, and the development of a velamen is quite widespread, if sporadic (von Guttenberg 1968 for literature).

Shoot apical meristem. See Hagemann (1967), Kaplan and Cooke (1997) and Bowman and Eshed (2000) for the basic organisation of the apical meristem. In angiosperms the apex of the stem is covered by layers of cells, the tunica, in which cell division in predominantly anticlinal direction, that surround the corpus, a central mass of cells in which the plane of cell division is not so restrained (this construction pattern is responsible for the distinctive variegation patterns in many taxa - see Tilney-Basset 1986). Tunica-corpus construction may be a synapomorphy for angiosperms (and a parallelism with Araucaria and Gnetales - see Johnson 1951, Fagerlind 1954, and Pohlheim 1971 for gymnosperm meristems; Gifford 1954 for details of tunica in angiosperms). Although details of organisation of the shoot apical meristem are known for relatively few plants, the numbers of corpus layers may be of some systematic interest (Bowman & Eshed 2000). Monocots examined have a primary thickening meristem (Rudall 1991 for a summary). There is considerable variation in the size of apical meristems, but those of Cactaceae seem substantially wider than the rest, being 400-1500 µm across, while in other angiosperms, including palms, they are <200(-300) µm across (Gifford 1954, see also references in Gifford & Corson 1971). I do not know of any more recent survey of meristem size.

Iqbal (1995) provides an entry into the literature on plant cambia, or lateral meristems (secondary thickening meristems), and the cells that differentiate from them. There are two main kinds of cambia, cork cambia and vascular cambia. The products of cork cambium (or phellogen), are cork or phellem to the outside and a little phelloderm to the inside; together the three make up the periderm. There is variation in the tissues produced as the cork cambium differentiates, when the cambium is first produced (see e.g. Möller 1882; Esau 1965; Edwards & Donoghue 2006), and, perhaps most importantly from a systematic point of view (Weiss 1890), where it is first initiated. Cork cambium occurs widely in broad-leaved angiosperms and in gymnosperms. In the stem it tends to be either superficial, the common condition in angiosperms, being initiated at or immediately below the epidermis (I have not distinguished between these positions), or deep-seated, being initiated deep in the cortex, inside the pericycle, or even in the phloem (for still invaluable surveys, see Möller 1882 [much information on gymnosperms]; Douliot 1889, also Kuhla 1987; Waisel 1995; etc.). Initiation in the cortex is also quite common and there is of course no absolute distinction to be made between the superficial and deep-seated positions. There may be a connection between the position of initiation of the cork cambium and the development of an endodermis (Arber 1925 and references). Czaja (1934) discusses the cork of several taxa that have cork wings on the young stem. Etagenkork, suberised cells that are not produced by a cork cambium, are also to be found in some monocots. Polyderm is a distinctive cambial derivative that is rather infrequently found, but whose distribution often seems to be of systematic significance (Mylius 1913 for a survey); the cambial cells cut off alternating layers of cork and endodermal cells.

In broad-leaved angiosperms, cork cambium position in the stem and root of the same plant commonly differs, the cork cambium in the roots usually being deep-seated (see Waisel & Liphschitz 1975; Esau 1977; Fahn 1990). However, in plants with aerial roots of various kinds the cork cambium may be superficial (von Guttenberg 1968); thus Clusieae, alone of Clusiaceae, are often epiphytic and have superficial cork cambium in their roots (but this is not found in Clusiella, a Clusia kook-alike that is in Kielmeyeroideae. In monocots the cork cambium in the root, when developed, is usually to be found just under the exodermis, i.e. in a superficial position (Philipp 1923; Arber 1925). However, specific information on the initiation of cork cambium in the roots is hard to come by, especially in monocots.

Whether or not a vascular cambium develops, and the nature of the products of that cambium, are is of considerable systematic interest. Vascular bundles in which cambium does not develop are described as being closed; if a cambium develops, normally at the interface between the phloem and xylem, they are open. Climbers in particular show anomalous patterns of secondary thickening, with the number of cambia, and the nature of the tissue they produce, varying. In non-climbing Caryophyllales such as Phytolaccaceae successive cambia develop. Carlquist (2004) studied the secondary tissue of Nyctaginaceae in detail, and suggested that there, at least, the use of the term "included phloem" was inappropriate; several taxa that had been described as having included phloem in fact have successive cambia. In successive cambia, new cambia develop to the inside of the cortex and produce conjunctive tisse, rays and a vascular cambia, which in turn generates phloem externally and xylem internally, hence the phloem is not completely included or enclosed by xylem tissue. Here the term included phloem is used to describe the situation when phloem islands cut off internally by a vascular cambium are completely surronded by xylem cut off by that same cambium, and internal phloem the situation where phloem is found internally to the xylem in the primary stem (this latter condition is commonly associated with bicollateral vascular bundles in the leaves). The terms "interxylary phloem" amd "intraxylary phloem" are confusing. A normal vascular cambium is absent in monocots, although there may be traces of cambial activity (Stant 1970 and references). In some monocots a distinctive cambium cuts off tissue to the inside in which separate vascular bundles embedded in ground tissue differentiate; this is scattered in Asparagales (Rudall 1995b for records and literature) and has recently been reported from Eriocaulaceae (Poales: Scatena et al. 2005). There is no vascular cambium in monocot roots (Arber 1925).

Data on cambium storying are taken mostly from Carlquist (1988b); if not mentioned, the cambium is unstoried. One or both the kinds of cambial initials (ray, fusiform) may be storied, but only rarely are ray initials alone storied, as perhaps in Lauraceae (Paul van Rijckevorsel [pers. comm.] noted problems with this character).

For information on the nature of the vessel element perforation plate in broad-leaved angiosperms, see Bierhorst and Zamora (1965), for monocots, see Wagner (1977, sampling needs to be extended/newer literature integrated). There has been discussion whether vessels in monocots and in other angiosperms arose independently (Cheadle 1953), although this seems unlikely; the distribution of vessels in the monocot plant body may be of systematic interest (Cheadle 1944; Wagner 1977). There is much information in the literature of the nature of scalariform perforation plates, e.g. the number of bars, and in the past the nature of the vessel perforations has been treated of very great phylogenetic importance, with the trend running tracheids only -> scalariform perforation plates with many -> few bars -> simple perforation plates; the cells also became shorter and wider and the end wall less oblique (e.g. Bailey 1944, 1954). I have not tried to fit all this information to the trees, and certainly the evolutionary significance of this variation is less than was thought, even if the Baileyan trends are more or less evident in the fossil record (Wheeler & Baas 1991). Pit membrane remnants are turning out to be quite common in broad-leaved angiosperms (e.g. Aextoxicaceae, Bruniaceae, Hydrangeaceae, families near Ericaceae, Schisandraceae, Quintiniaceae, Rousseaceae), but the functional and systematic significance of this feature is unclear (e.g. see Schneider & Carlquist 2003, 2004a; Carlquist & Schneider 2004).

Some information is included about the presence of vascular tracheids and wood fibers, and the presence and distribution of xylem parenchyma (see Hess 1950), but these features are not treated for all families.

Pits. Vascular pits are in general quite variable in morphology and arrangement, although the distinction between some of the pit "types", such as bordered and non-bordered pits is arbitrary. The nature of intervascular pitting may be of systematic interest, but I have not included many details of this character. Hacke et al. (2004) describe how the torus-margo pit membranes of gymnosperms work; their absence may be an angiosperm apomorphy (see also Sperry & Hacke 2004). However, a torus, sometimes lignified, occurs occasionally in angiosperms (Coleman et al. 2004), and again its distribution may be of systematic interest. Vestured pits have close and minute sculpturing of the surface of the cell wall in the pit, and their distribution is of systematic interest (for surveys, see Jansen et al. 1998, 2001). Nearly all taxa with vestured pits have vessels with simple perforation plates, and such pits may promote the functioning of these vessels (Jansen et al. 2003).

Reaction wood. The amount and to a certain extent the anatomy of the xylem produced at the branch-stem junction, the reaction wood, varies, although in some taxa it may not be produced at all. Although it is often suggested that angiosperms and gymnosperms differ in the nature of their reaction wood, angiosperms producing cellulose-rich tension wood and gymnosperms lignin rich compression wood, a summary of the information available for angiosperms (Höster & Liese 1966) suggests that this is a gross oversimplification (see also Westing 1965, 1968 and Timell 1986 for gymnosperms).

Wood fluorescence. Data are taken largely from a survey of some 10,600 species summarised in Avella et al. (1988). I am unclear as to the significance of this character.

Unlignified cell wall fluorescence is caused by the presence of ferulic and/or coumaric acids. Unlignified walls of most angiosperms do not fluoresce, and only those taxa with fluorescence are mentioned (data largely from Hartley & Harris 1981). Other information on the composition of primary cell walls, especially in monocots, is given by Smith and Harris (1999) and Harris (2000).

Details of vascular ray "types" are rarely mentioned, partly because it is difficult to reconcile the definitions of the terms as they have been used over the years (see Kribs 1935, 1959), partly because of the typology involved, and partly because it can be difficult even to define what a ray is (Carlquist 2003); in general systematically interesting information in this feature is probably best recorded in descriptive terms, i.e., describing the morphology of the cells involved. The short discussion by Chalk (1983) is particularly useful. Particularly distinctive rays are found in some Malvaceae (e.g. Chattaway 1933a, b). Ray width in both xylem and phloem may vary considerably. Broad phloem and xylem rays are a valuable field character for identifying families, and when the phloem rays are wide, tangential cuts of the phloem show a net-like arrangement of the paler-colored, anastomosing rays (Keller 1996). More precise indication of ray width can provide systematically interesting information.

Phloem. In gymnosperms sieve cells (Strasburger cells) are involved in carbohydrate transfer; these are nucleated and have endoplasmic reticulum traversing the sieve pores (e.g. Iqbal 1995). In angiosperms sieve tubes carry out the same function; they lack nuclei but are closely associated with nucleated companion cells that are derived from the same immediate mother cell. There is little variation in sieve tube morphology easily usable at higher levels, although Zahur (1956) described substantial variation in many characters, albeit it seems to be overlapping (Iqbal & Zahur 1995 list trends of specialization). Zahur (1956) found a weak correlation between presence of secondary septae and sympetalous families, but his data show variation within orders. Turgeon et al. (2001; see also Gamalei 1991) summarise variation in companion cell anatomy and delimit a number of "types". One of these types, intermediary cells, is characterized by having numerous plasmodesmata that branch in the outer part of the walls adjacent to the bundle sheath cells; there seems to be a correlation between the presence of intermediary cells (see Lamiales) and the presence of stachyose in the translocate. However, whether or not such characters are more or less constant within families is not obvious.

Transfer cells, metabolically very active cells with more or less labyrinthine intrusions of the walls, occur in various parts of the plant (Gunning et al. 1970; Pate & Gunning 1972), e.g. the heads of the glandular hairs in insectivorous plants, associated with phloem in vascular bundles (they are one type of companion cell), especially in the xylem and at the nodes, and in seeds (see below).

Stratified phloem, secondary phloem in which there are bands of fibres alternating with ordinary ploem tissue, is developed in groups like Malvales, Annonaceae, etc. It is easy to recognise using a hand lens, or even with the naked eye, in phloroglucinol-stained hand sections. Plants with such phloem typically have bark that can be pulled off in long strips.

An important character is the structure and composition of sieve tube plastids. Data on the ultrastructural morphology of the starch and protein in the plastids are taken from the work of Behnke (e.g. 1969, 1981b, 1989, 1991a, 2000, 2001). The basic condition for angiosperm plastids is to include starch grains alone, but nearly all monocots have cuneate protein crystals alone and core Caryophyllales have peripheral protein fibres. There are many other less striking variants with various combinations of starch and protein, and these may characterise small clades such as [Erythroxylaceae + Rhizophoraceae], whilst the absence of both starch and protein is common in parasitic groups, but also in Rosales, [Malpighiaceae + Elatinaceae], etc. The sizes of the starch grains and fine details of the protein crystals in the plastids may provide still more evidence of relationships, although I have not always included this information. Note that Pistia is the only monocot known to have starch plastids (Behnke 1995)!

P-proteins are proteins in mature sieve elements of angiosperms that vary in origin (nuleus or elsewhere) and appearance aggregated [non-dispersed] or not) and also behaviour, e.g. changing their morphology when the turgor of the sieve tube changes (Knoblauch et al. 2001). Information on this character, varying mostly in broad-leaved angiosperms, was obtained largely from Behnke (1981a, 1991b); P-proteins are absent from at least some monocots, e.g. Poaceae, but they are present in e.g. Iridaceae (Tóth & Sjölund 1994; Sabnis & Sabnis 1995).

In the primary stem vascular system there can be separate bundles (an eusetele), a (pseudo)siphonostele, with the vascular tissue forming an apparently complete cylinder, or a complete cylinder. However, from Benzing's work (Benzing 1967b) there seems to be no sharp difference between siphonosteles and eusteles in angiosperms. Moreover, the use of such terms may be inappropriate when describing the vascular system there (Benzing 1967a; Namboodiri & Beck 1968c; Howard 1979a), which can often be interpreted as a series of sympodially branching bundles; variation within families may be quite extensive (Jensen 1968). Finally, interpretation of the primary vasculature in terms of axial or foliar theories yields very different results (Al-Turki et al. 2003) - and of course the theories suggest very different evolutionary pathways for the plant body. See Schmid (1982; also Namboodiri & Beck 1968a, b; Beck et al. 1982) for a discussion of stelar variation and terminology in both angiosperms and gymnosperms. Beck et al. (1982) suggest that there is a correlation in broad leaved angiosperms between closed primary stelar patterns and distichous or opposite phyllotaxy and between open patterns and spiral phyllotaxis.

In monocot stems the vascular bundles are usually scattered, and I mention them only when they are in rings. In general, cauline vascular bundles are collateral, with xylem to the center of the stem and phloem to the outide. Variants of this are interesting, and are mentioned when they occcur; this variation is often reflected in the general arrangement of tissue types in the vascular bundles of the petiole and midrib. In monocot rhizomes in particular, but not in the annual stems, the scattered bundles are frequently amphivasal, i.e., they consist of phloem surrounded by xylem (Jeffrey 1917; Arber 1925).

For nodal anatomy see especially Sinnott (1914), Dormer (1946), Marsden and Bailey (1955), Bailey (1956), Balfour and Philipson (1962), Cutler and Gregory (1998), Howard (1970, 1974), Beck et al. (1982), and Metcalfe (1987) for information. Note that the behavior of protoxylem strands in the young stem may not correlate with node type (e.g. Beck 1962; Kumari 1963; Benzing 1967a, b; Gibson 1994), and treating flowering plants as if they had a modified siphonostele (see above) may obscure information (Keating 2000). In particular, although I refer to "leaf gaps" in these pages, Beck et al. (1982) in particular argue that this is inappropriate when talking about the steles of seed plants. Indeed, when looking at nodal anatomy in stems with some secondary thickening, it may be very difficult to detect nodes in which there are separate traces from adjacent sympodia. Also, it should be remembered that the stem, node and leaf form a continuum (see especially Howard 1974, 1979a), and there can be differences in nodal vascularisation when cotyledons and foliage leaves are compared (e.g. Stone 1970). In the descriptions I use formulae like 1:1, 5:5, etc.; the first number is the number of gaps, the second is the number of traces or separate vascular bundles leaving the gaps. These gaps are often recognizable as interruptions in the vascular cylinder formed by the early stages of secondary thickening. Split laterals occur in some taxa with opposite leaves and are single traces that send branches to both leaves, while the related flank-bridges and variants are found in some Dipsacales and Rubiaceae (e.g. Neubauer 1982), etc. The cotyledonary node often has 1:2 nodal anatomy (Thomas 1907), whatever the nodal anatomy of the later leaves. However, there can be quite a lot of variation at the cotyledonary node; it may be multilacunar in Juglandaceae, or 4:3 in some Magnoliaceae (with split laterals!, e.g. Sugiyama 1976a), and the basic nodal condition for seed plants may be 1:1 (Kumari 1963). Although many asterids, especially Lamiales, etc., are commonly reported as having 1:1 nodes, Neubauer (e.g. 1977, 1978) emphasized that there the single trace often divided immediately into three or more (e.g. see Linderniaceae, Scrophulariaceae, also Cordia [Boraginaceae]); nodal anatomy may repay attention in Lamiales (also Melastomataceae, etc.).

Leaves that have stipules often have blades with serrate margins and are not often unilacunar (Celastraceae are an important exception to this latter feature, although their stipules can be the most evanescent of structures). Rosaceae tend to have 3:3 or more complex nodes, and are generally stipulate; the estipulate Spiraea has 1:1 nodes. Many plants with 1:1 nodes have leaves with entire margins (Sinnott & Bailey 1914). Monocots have broad bases into which several bundles proceed, but Sinnott and Bailey (1914) noted that in the petiolate Potamogeton and Smilax there was a tendency for three main bundles to enter the leaf base. Gross patterns of nodal anatomy are often easy to confirm in the field using a razor blade or sharp knife and a hand lens, and so can help in identification. Thus Celastraceae have 1:1 modes, different from most Malpighiales, while both Perrottetia and Bhesa, recently removed from Celastraceae (and now in Huerteales and Malpighiales respectively), have more complex nodes - and more obvious stipules, especially in Bhesa!

In monocots a number of vascular bundles proceed in to the leaf, and details of nodal vasculature are not generally recorded. In aquatic angiosperms like Nymphaeales nodal anatomy is very complex, with a vascular plexus forming at the nodes and the internodal vascular tissue being very distinctive, even if ultimately the vasculature of the leaves can be interpreted in conventional angiospermous terms (cf. Schneider 1980; Moseley et al. 1984). Gymnosperm nodal anatomy is also described in terms of gaps and traces, although at the very least the behaviour of the traces can be unexpected from an angiosperm perspective, as with the girdling traces of the foliage leaves in Cycadales.

There are often lignified fibers immediately external/peripheral to the phloem in the stem, the pericyclic sheath. This sheath is sometimes absent, or it may be made up of sclereids. Gibson (1993) discussed the need to distinguish between fibers that are derived from phloic tissue and those that are extra-phloic in origin (see also Blyth 1958). Although simple presence/absence and nature (if present) of this sheath may be of systematic interest, as in Schlegliaceae, Dipsacales, etc., distinguishing between different origins of the sheath may also be of value.

The endodermis consists of one or a few layers of cells that surrounds the central stele and sometimes also vascular tissue elsewhere. In the root the endodermis is often marked by a single layer of cells with casparian bands, but this is uncommon in the stem and even more so in the leaf, although it may occur there. In the stem a ring of starch-containing cells (the starch sheath) may be found in this position (Van Fleet 1961). Details of the distribution and nature of cauline and foliar endodermes in particular may be of systematic significance, see e.g. Alseuosmiaceae.

Cortical and medullary bundles, with vascular bundles in the cortex and pith respectively, are mentioned only when they occur, see Col (1904) for an early discussion and summary of information.

Petiole anatomy has been recorded rather sporadically. Radford et al. (1974) describe and name some of the main variants, although other sets of terms for describing petiole vasculature have also been proposed (e.g. Watari 1939; see also Howard 1979b). Col (1904) is a source for information in several groups. Families like Clusiaceae show great variation in this feature, and it is usually of systematic significance at levels lower than those emphasized here. However, Kim et al. (2003) have recently drawn attention to variation in how much adaxial surface is evident in petioles; when there is none the petiole bundles form a circle, or there is a single annular bundle, in both cases without any adaxial bundles (the "flange bundles" of the descriptions). Vascular bundles in the petioles of monocots are usually scattered, although sometimes they form a ring; only the latter feature is recorded.

Various tissues in the plant contain distinctive secretions. Latex is generally secreted in laticifers, rarely in isolated cells. The former may be unarticulated, branched or unbranched but not anastomosing syncytia that originate from single cells in the embryo. Articulated laticifers, generally associated with the phloem, are often anastomosing tubes that develop from cells whose common end walls have broken down. Gregory and Baas (1989; see also Dickinson 2000) provide a useful survey of mucilage cells, while Matthews and Endress (2006) distinguish a distinctive type of mucilage cell with thickened inner periclinal wall and distinct sytoplasm; they catalogue the distributions of this and of other kinds of mucilage cells.

Leaf blades containing ethereal oil cells may be punctate when a light is shone through the leaf. In glands of taxa like Rutaceae and Myrtaceae material is secreted into schizogenous cavities; truly lysigenous glands may not occur in plants, reports of such structures seeming to refer to artifacts of preparation (Turner et al. 1998; Turner 1999). Cells containing ethereal oils and those with mucilage are similar in their early development, as in basal angiosperms (Bakker 1992). The main differences between the two is that oil cells often have a suberised layer between cellulose layers and plastids secrete the oil, while mucilage cells lack the suberised layer and mucilage is usually secreted in golgi vesicles.

Sclereids (sclerenchymatous idioblasts) found free in the cortex, the mesophyll of the lamina, or elsewhere are variable in morphology; they are mentioned when they are common in a family (Rao 1991 for a survey). Sclereids occuring in the pith are rarely mentioned. Sclereids intergrade with fibers, very elongated and unbranched lignified cells. Distinctive sclereids of various kinds, including tracheoidal cells, may be associated with vein endings; such sclereids are rarely mentioned below because they are of significance at lower levels. Again, leaf blades with sclereids may be punctate when a light is shone through the leaf, and even fibrous tissue is often evident, either in such transmitted light or in surface view.

Cristarque cells are distinctive cells with U-shaped lignification and with calcium oxalate crystals or druses. They are mentioned only when they have been reported.

Calcium oxalate may crystallise in distinctive forms such as raphides (needle-like crystals especially common in monocots), styloids (narrowly oblong crystals), crystal sand (very small crystals that appear granular), or single rhomboid crystals little longer than broad; crystals of any sort are rarely entirely absent. All these conditions are mentioned. Raphides can sometimes be seen in herbarium material using a dissecting microscope if the tissue containing them is cut; the raphide sacs are then evident as small white patches. Calcium oxalate crystals are usually in the form of druses, irregular aggregations of more or less radially arranged crystals, but then their presence is not mentioned. Many data are taken from Metcalfe and Chalk (1950); Prychid and Rudall (1999, 2000) provide surveys for monocots, Lersten and Horner (2005) discuss styloid morphology and distribution, and Franceschi and Horner (1980) review the subject from a more physiological perspective. Calcium oxalate may crystallise as whewellite [CaC2O4.H2O] or weddellite [CaC2O4.2H2O], the latter being the less common form (Monje & Baran 2002), but I do not know what the general distribtutions of these forms are (see Cactaceae). Cystoliths are moderately common and their distribution is quite often of systematic interest. They are concretions of calcium oxalate and/or calcium carbonate developing on an intrusion of the cell wall; the cells containing cystoliths are known as lithocysts. Cystoliths can sometimes be seen in the leaf blade when it is held up to the light, as in Acanthaceae and Urticaceae.

Crystalline Silica (SiO2) is sometimes present, either in the wood of families like Chrysobalanaceae, or in the leaf, as many Poales. It may be present as sand, or as larger, more organised bodies of distinctive shapes. Presence or absence of silica bodies, where they are deposited in the plant, and the morphology of the silica bodies provide especially valuable characters in the monocots (see Prychid et al. 2003b for a summary). Ma and Takahashi (2002) summarize the literature on silicon concentrations in plants. Although most broad-leaved angiosperms and non-commelinid monocots have low concentrations while commelinids have high concentrations, some commelinids, including members of Commelinales and Juncaceae sampled, also have low concentrations, but these groups also lack SiO2 bodies.

For a useful general discussion of variation in the leaf surface, see Wilkinson (1979). Hydathodes, extra-floral nectaries, domatia, and even lenticels may be found on the leaf surface, although they rarely mentioned in these pages because their variation is mostly at a level lower than is the focus here. Hydathodes are particularly well supplied with tracheids, not with phloem; extra-floral nectaries show the reverse relationship (for the expression of the CRABS CLAW gene in both floral and extrafloral nectaries in core eudicots, see Lee et al. 2004). Extra-floral nectaries may of course be found on various parts of the plant - stem, leaf, the outside surface of the calyx - and a distinction is sometimes made between such nectaries that are intimately involved in pollination, as in those of the cyathia of Euphorbia (nuptial nectaries) and extra-nuptial nectaries that have nothing immediately to do with pollination, but may attract ants that help protect the plant against herbivores or destroy young stems of vines that would otherwise use the plant as support. For a survey of extrafloral nectaries, see Elias (1983).

Epidermis. Epidermal cells may contain silica bodies (see above) or their walls may be lignified. In many monocots there are bulliform cells, large, thin-walled cells on the adaxial epidermis, that help cause curling and uncurling of the leaf blade as their tugor changes; subepidermal cells may be involved as well (Löv 1926; Linsbauer 1930). Epidermal cell shape is usually of little interest at the genus level and above.

Stomatal type. The stomatal types mentioned are those delimited by Van Cotthem (1970: slightly modified by Baranova 1983, see also Baranova 1987, Wilkinson 1979), however, there is much disagreement in the literature over stomatal morphology, even that of smaller families. Stomatal "type" is variable, even when definable, but further confusing the issue is the tendency for each author to use a different set of terms or to interpret a particular configuration of cells in a different way (Carpenter 2005 recently introduced nine new types in an exhaustive study of stomatal variation in the ANITA grade, but the forty four types recognised by Patel in 1979 [and largely ignored since] is probably the record). As Payne (1970) observed when introducing his helicocytic and allocytic stomatal types, what may seem to be a distinct stomatal type to one person can be considered a mere modification of an existing type to another person. Baranova (1987) is a useful guide through the nomenclatural chaos, as she noted "The distinctions among these types, although conceptually useful, are often difficult to draw. We should not let the necessity to categorize the diversity among stomatotypes obscure the fact that varition is continuous." (Baranova 1987: 64). The patterning of cells around hair bases, epidermal oil cells, etc., can also be described using the same set of terms, even if the idea that all these structures are homologous seems unlikely (cf. Carpenter 2006). Stebbins and Khush (1961) provide an early stomatal survey for monocots, although Tomlinson (1974) qualified their account, emphasising the importance of stomatal development (see also Rudall 2000). Note that stomata with different ontogenies occur in Poales as circumscribed here (Tomlinson 1969). Illustrations in Conover (1991) are useful for ascertaining stomatal morphology of some monocot groups. Inclusions in the guard cells may be of systematic interest (see Tiemann 1988).

Stomatal orientation also varies. In monocots with parallel venation (see below), the stomata are usually oriented with the long axes of their apertures parallel to that of the blade, however, in those monocots with reticulate venation orientation is often random (Conover 1983). Elsewhere, random orientation is common, although there are exceptions. Butterfass (1987) summarises information on taxa that have tranversely-oriented stomata.

Cuticle waxes. Data are taken from Barthlott and his coworkers (e.g. Fehrenbach & Barthlott 1988; Ditsch & Barthlott 1994, 1997; Barthlott et al. 2003). Barthlott et al. (1998) standardised the terms used in describing wax crystals, and noted that unoriented platelets, often with more or less irregular margins, are common; indeed, their distribution at and above the familial level seems of little systematic significance. Below I usually mention only other forms assumed by the waxes and their chemical composition (for which, see Meusel et al. 1999; Barthlott et al. 2003). In general, wax crystalloid type is more important in monocots (e.g. Barthlott & Fröhlich 1983), but even there the distinctive types are often disconcertingly sporadic in their occurence. Wax crystals, cuticular folds (Barthlott & Ehler 1977) and also trichomes (Barthlott et al. 2003) rarely occur together on the one plant, with wax crystals predominating in "basal" angiosperms (but they seem to provide little information of systematic interest in the ANITA grade) and monocots, while a well-developed indumentum, and so conversely poorly-developed wax crystals, predominate for example in Lamiales, Solanales, and Asterales.

GENERAL MORPHOLOGY

Still extremely useful are compendia such as those of Arber (1925) for monocots, and Arber (1920) and Sculthorpe (1967) for aquatic plants, Goebel (1931, 1932) and Troll (1935-1971); Bell (1991) provides an attractive short survey of plant morphology. Vegetative characters have been emphasised where possible, especially in the thumb-nail sketches of families, and here Hallé, Oldeman & Tomlinson (1978), Keller (1996), van Balgooy (1997, 1998, 2001) and Schatz (2001) have been valuable sources of information. However, many vegetative characters are best used in a more local context, identifying members of the local flora; for identification using vegetative characters, see e.g. Keller (1996), van Balgooy (1997, 1998, 2001), Schatz (2001), Rejmánek and Brewer (2001) and Spichiger et al. (2005).

The vegetative body

Hair type. Only particularly distinctive hair types are mentioned, and these rarely suggest relationships between families, although they are quite useful at rather lower levels. There is the same problem here as with stomatal morphology, etc.: "Because of the intergradations between trichome types, and also in consequence of the use of imprecise and varied morphological and histological terms, it is often difficult, from the voluminous published descriptions that exist, to determine exactly what type of trichome is being discussed, unless there is an illustration." (Theobald et al. 1979, p. 41). For terminology, see Payne (1978), Hewson (1988), Stearn (1992), etc.

The vegetative bud can be perulate (scaly) or "naked", that is, lacking morphologically modified leaves enclosing the apex. Information on this character is only sometimes mentioned.

Branching. Keller (1996) distinguished between immediate branching, with branches coming from the currrent flush, and delayed branching, with branches developed on the previous flush. This is close to the distinction between prolepsis and syllepsis, whether branches develop from the shoot with (prolepsis) or without (syllepsis) evident rest. However, attributing such features to taxa can be difficult, since branching behavior on sucker shoots may differ from that elsewhere on the plant. Immediate branching is relatively rare in temperate taxa, but may occur in the local representatives of otherwise largely tropical families that commonly have such branching (e.g. Lauraceae). Buds developing from the axils of the last pair of leaves of the previous flush (e.g. Clusia, Garcinia) present problems for such distinctions, since they may grow out simultaneously with the elongation of the main axis, but after a period of rest. Branching in vines and annuals may also depart from a family norm. Although this character is mentioned quite frequently in the characterisations, sampling is poor.

Prophyll number and orientation. Prophylls are usually either paired and lateral (the basal condition) or single and adaxial, although Piperaceae in particular seem to show variation in this. Whether or not prophylls are basal on the branch is often, but not always, associated with whether branches are proleptic or sylleptic; in proleptic branches the prophylls are usually basal, while in sylleptic branches they have a well-developed internode, the hypopodium, below them. Exceptions have both biological and systematic significance. Thus most Ulmaceae, Cannabaceae, etc., appear to have basal prophylls whatever the dynamics of shoot growth, and buds in the axils of these prophylls may be very conspicuous; inflorescences are often axillary paired or paired and borne at the bottom of branches (they come from the axils of the prophyllar buds). Taxa with axillary thorns or branch tendrils - these are usually sylleptic - may also show distinctive patterns of variation. Prophylls that are not basal on the thorn (e.g. Gleditsia) are often associated with the development of serial buds in the axils of leaves of the main stem (these provide replacement meristems for the main axillary shoot = lignified and dead thorn), while thorns with basal prophylls (e.g. Crataegus) have functional buds in the axils of these prophylls and lack serial buds in the axils of the leaves of the main stem. Note that bracteoles (see below) are simply the prophylls of the flowering shoot. Especially in monocots, there is extensive discussion as to what prophylls "are" - one leaf, two connate leaves?; Blaser (1944) has summarised the often exasperating early literature on the topic.

Leaf insertion. I record the insertion of leaves on flowering branches, not on non-flowering axes (where these are distinct); the variations are spiral, 2-ranked (distichous), 3-ranked (tristichous) and opposite (including bijugate and whorled). Mabberley (1997), too, makes the important distinction between spiral and distichous (see also Corner 1946a), which are too often lumped together and ambiguously referred to as 'alternate'. For systematically informative variation in leaf insertion on the main stem/trunk when leaves on branches are invariant in their insertion, see Johnson (1993); note that this variation does not affect architectural models. I make no distinction between opposite and whorled leaf insertion in the family characterisations since this commonly varies within a family or genus; most families in which opposite leaf insertion is common also have taxa with whorled leaves. For the control of phyllotaxis, see Reinhardt et al. (2003), for a good brief discussion on phyllotaxy and its measurement, see Rutishauser and Peisl (2001).

For details of general leaf development, whether the lamina develops from the leaf base (monocots) or leaf apex (broad-leaved angiosperms), see Kaplan (1973), Bharathan (1996), Rudall and Buzgo (2002) and Nardmann et al. (2004); Sinha (1999) reviews leaf development in angiosperms. The terete, unifacial blades with stomata all over the surface that are found in monocots may result from the development of the unifacial "tip" or "Vorläuferspitze" of the developing leaf (e.g. Arber 1925; Troll 1955; Troll & Meyer 1955; Kaplan 1975 - Oxypolis [Apiaceae] and monocots that are compared in this last paper would seem rather distant), or from the middle portion of a bifacial leaf (see illustrations in Linder & Caddick 2001, for which see also for a summary of the literature). Nevertheless, patterns of leaf development may vary at high levels. "Unifacial" leaves in broad-leaved angiosperms may have stomata on both sides of the blade (amphistomatic), but the blade is still dorsi-ventrally flattened, i.e., they are isobifacial. In some monocots, such leaves are often equitant at the base; they, too, are isobifacial, but they are laterally flattened. They appear to represent a normal bifacial dorsiventral blade that has folded and become connate adaxially, or they could be developed from a "midrib". Indeed, much more work is needed on this suite of characters: the development of an isobifacial monocot leaf such as are found in Acorus can be explained in quite different ways - is it a hyperphyll, and so equivalent to the broad angiosperm leaf, or does it originate from an intermediate zone between hyperphyll and hypophyll (Rudall & Buzgo 2002)? Similarly, the "blade" of Hosta and Orontium may not be equivalent structures in other than a functional sense (Troll 1955), with the ligule (see below) in at least some cases in a monocot demarcating the Vorläuferspitze from the rest of the leaf - note that in Zamioculcas there is a ligule very near the base of the petiole...

Leaf size. Ecologists in particular can find it useful to divide the size of the leaf or leaflet blade into a number of classes (Raunkiaer 1934; Webb 1959). Rather small, narrow leaves with very strongly revolute margins, ericoid leaves, are quite common in taxa that grow in warm and dry climates.

Leaf type. The main distinction is between simple and compound leaves. However, this distinction may not be so clear-cut as it seems. Bharanthan et al. (2002) show how in many leaves dissection in general - whether resulting in simple lobing, or in fully compound leaves - is associated with expression of the KNOXI gene. Normally it is not expressed after leaf initiation on the apical meristem. Some simple leaves may show KNOXI expression early, but lobing is obliterated by inner blade growth; all simple leaves are not identical. Many taxa that commonly have compound leaves have some members with apparently simple leaves; there, however, the blade may be joined to the petiole by an articulation, i.e. the leaves are reslly unifoliolate, not simple. A number of taxa with pinnate leaves also have palmate leaves, however, Kim et al. (2003) have recently made a distinction between peltately palmate and non-peltately palmate leaves. In peltately palmate leaves the petiole bundle is strictly annular (or the bundles form a circle), there being no adaxial surface, whereas in non-peltately palmate and pinnate leaves the petiole bundle is more or less obviously dorsiventral, the adaxial surface being evident. Monocots, with the exception of a few Araceae and Dioscoreaceae, lack truly compound leaves in which the the leaflets arise from separate primordia along the leaf, the activity of the blastozone having become restricted to those areas that later develop into leaflets (see Gunawardena & Dengler 2006 for a useful review). Apparently compound leaves such as those common in Arecaceae, as well as fenestrate leaves, as in Araceae and Aponogetonaceae, are caused by localized cell death (e.g. Gunawardena et al. 2004).

Lamina ptyxis (vernation) - terms and some data are taken from Cullen (1978) and Keller (1996: note that his convolute is Cullen's supervolute). In compound leaves, ptyxis refers to the folding of the leaflets. The sampling of this character is poor, the delimitation of the states needs attention, and variation may turn out to be rather too much for comfort when used at the level of this site. When leaves remain rolled up as they elongate, the lower surface in particular of the fully expanded blades may have faint but distinct longitudinal lines; these often help in familial or generic identification of sterile material. Three generalisations can be made about ptyxis type: first, in water plants with floating, peltate or cordate leaf blades, involute ptyxis is very common, second, the leaflets of compound leaves tend to be conduplicate, and third, ptyxis in monocots is supervolute or supervolute-curved.

Lamina margin. In the category "toothed" are all margins that are serrate or dentate, even minutely so; the margin may even appear to be smooth, but close inspection will show small, more or less regularly spaced dots or glands. A number of tooth morphologies have been described for the leaves of broad-leaved angiosperms (Hickey & Wolf 1975), but these are difficult to recognise on simple inspection of a leaf. Intermediates between these morphologies are known, as in Fagales (Hickey & Taylor 1991), where two "types" and their intermediates may occur on the one leaf. However, tooth morphology may yield information of systematic significance if analysed more carefully than it has been treated here. Note that nearly all families described as having leaves with teeth have some members with entire margins. In monocots, leaf teeth are never glandular but can be more or less spiny.

Punctations, often "glands" of various kinds that are embedded in the mesophyll, are evident when the leaf is viewed in transmitted light, provide evidence of these anatomical features and are useful in identification. Note that magnoliids with ethereal oils tend to have ± punctate laminas (e.g. van Balgooy 1997), as well as the text-book examples of Rutaceae and Myrtaceae, where the glands are often much larger and more organised. Other taxa secreting oils, tanniniferous substances, etc., or with cystoliths or sclereids (see below) may also be punctate, while punctations may also be evident in the calyx, corolla, etc. Punctations are mentioned only when they are conspicuous.

Lamina venation. A distinction is drawn between leaves in which two or more major veins leave the midrib at or almost at the base of the lamina (palmate), those in which these veins leave the midrib along its length (pinnate), and those in which there is no distinction between a midrib and secondary veins, all veins being similar, parallel, and proceeding to the apex (parallel, obviously especially common in monocots); further details of venation are rarely given. Klucking (1986 onwards) provides a comprehensive survey of leaf venation and Christophel and Hyland (1993) is another useful source of information, however, the detail shown by Ettingshausen (e.g. 1858) has hardly been improved upon. Hickey and Wolfe (1975), Hickey (1979) and the Leaf Architecture Working Group (1999) define terms used in describing venation. Conover (1983) and Inamdar et al. (1983) discuss leaf morphology and architecture in broad-leaved monocots with reticulate venation; such leaves have evolved several times and are associated with particular ecological conditions (e.g. Givnish et al. 2004c). In many broad-leaved monocot taxa the parallel main veins are linked by well developed ladder-like cross veins. Nelson and Dengler (1997) surveyed the development of leaf venation. Unfortunately, sampling within broad-leaved angiosperms was largely restricted to Rosids, and within these, Arabidopsis (Brassicaceae) differed in having basipetal development of tertiary venation (it is normally acropetal in Rosids). Taxa such as Liriodendron (Magnoliaceae) have an intermediate pattern of initiation.

The leaf base is commonly sheathing in monocots; the sheath may be open, with free edges, or closed and continuous around the stem. It may bear distinctive, adaxial flanges or ligules at the very base or some way up the sheath; auricles, lateral projections of the petiole" or sheath/blade junction, also occur. The ligule in at least some cases in a monocot demarcates the Vorläuferspitze from the rest of the leaf (see above).

Stipules in broad-leaved angiosperms are more or less foliaceous, vascularised, often paired outgrowths at the base of the petiole or on the stem lateral to the petioles; they are a part of the leaf. They may be petiolar, clearly inserted on the petiole (and falling off when the leaf falls, if they have not fallen before), or cauline, largely inserted on the stem. Intrapetiolar stipules are inserted between the petiole and stem, sheathing stipules (= ochreae) surround the stem in a continuous tube, whether or not they are more or less scarious, as in Polygonum, or thick and robust, as in Coccoloba and some species of Fagraea, while interpetiolar stipules are single, often lobed stipules borne between opposing leaves. No attempt has been made to distinguish between "true" stipules and other structures; some things called stipules in families like Burseraceae and Sapindaceae are clearly modified colleters or leaflets, although when this is obvious the characterisations are qualified. Useful studies of stipules and stipuliform structures may be found in Weberling (e.g. 1968, 1970, and especially 2006 [largely a summary]) and Weberling and Leenhouts (1965).

Some suggest that stipules ex definitio do not occur in monocots. However, as has been pointed out by authors like Roth (1949) and Rudall and Buzgo (2002), the developmental origins of monocot ligules and at least some stipules of broad-leaved angiosperms seem not to be fundamentally different; both arise from adaxial cross meristems in the transition zone between hyperphyll and hypophyll. Although I have not used the term "stipule" in the monocot characterisations, it is probable that there are structures there that have at least as good a title to the name as some of the things called stipules in broad-leaved angiosperms.

Other aspects of leaf morphology are mentioned only occasionally. These include the presence of abaxial cavities or domatia, which tends to be a character varying at the level of species or sometimes genera.

Colleters are multicellular, more or less glandular but not vascularised structures secreting resin and/or mucilage and usually found in the axils of petioles or stipules - and quite often the calyx, too. The "intravaginal squamules" of monocots are included here, their only difference from colleters (a heterogeneous class of structures in the first place) seems to be that they are flattened (Doyle & Endress 2000). A useful survey is provided by Thomas (1991), but the distribution of colleters is definitely under-recorded. They may occur in other than the axillary position, e.g. several Apocynaceae, Rubiaceae, and Caricaceae.

The breeding system predominating in the family is often mentioned. Any sizeable family recorded as being monoecious or dioecious is rarely consistently so, conversely, families predominantly with "perfect" flowers (the common condition, and not mentioned when it occurs) nearly always have some taxa with other flower types. When staminate and carpellate flowers occur on the same inflorescence, the carpellate flowers are usually topologically basal; this is true of both monocots (e.g. Tomlinson 1982; Rudall 2003; Ambrose et al. 2005) and broad-leaved angiosperms (e.g. Krosnick et al. 2006). Exceptions in the latter group include Begonia (Begoniaceae) in which the carpellate flowers are produced in the cymose inflorescence after the staminate flowers, Ricinus (Euphorbiaceae) in which the carpellate flowers are in the top part of the inflorescence, and even the fossil Archaefrucuts, at least under some interpretations of its morphology (Friis et al. 2003b).

Inflorescence. For a generally accessible introduction to inflorescence morphology, which can be very complex and with a correspondingly complex and daunting terminology, see Weberling (1989, also Troll & Weberling 1989). Briggs and Johnson (1979) suggest a "more logical (but not theory-limited)" approach to inflorescence description and comparison, but it is difficult to apply to a study at this level, while Claßen-Bockhoff (2000) discusses previous attempts to describe and classify inflorescences (and suggests a general evolutionary schema for flowering plants herself). If one decomposes the basic growth processes that are likely to be going on in inflorescences, the variables are encompassed by thinking of an axis that is terminated by a flower, or not, that keeps on growing, or dies, and inflorescence bracts/prophylls that subtend branches or do not; the same goes on at all orders of branching. Add to that the degree of elongation of the internodes, and that is basically it (Kellogg 2000b).

Here I refer mainly to racemes or racemose (polytelic) inflorescences where the main axis does not terminate in a flower, the flowers being single in axils along the axis, and cymes of various kinds, when there is no extended main axis and all lateral axes are immediately terminated by flowers (for a discussion of the various kinds of monochasial cymes, see Buys & Hilger 2003). In a thyrse the axes main axis may or may not be terminated by a flower, but the branches are cymes, and in a panicle ("Rispe") the inflorescence is branched, there are no simple, bracteate cymes, but all axes are terminated by flowers (extreme polytely). Capitulae, corymbs and umbels are also mentioned in the characterisations below. Leins (2000) distinguished between indeterminate (open) and determinate (closed) inflorescences. The former included racemes, etc., as well as thyrses, rather confusingly defined as inflorescences in which the main and sometimes the lateral axes might not be terminated by flowers, while the latter included "Vorblatt-Inflorescenzen" (cymes), panicles, etc. [This is all very unsatisfactory.]

Inflorescences may be axillary and/or terminal, or they may be borne along the branches or even at the base of the trunk. A few taxa are geocarpous, with at least some inflorescences being underground (Kaul et al. 2000); the terms used to describe this and other similar behaviours are complex (Barker 2005). Epiphyllous inflorescences that arise from along the petiole or lamina, tend to be sporadic in occurence (see Stork 1956; Dickinson 1978), but cf. Dichapetalaceae and Phyllonomaceae + Hellwingiaceae.

Bracts and bracteoles are usually mentioned only when they are absent; the position of the latter is normally the same as that of the prophylls (see above), since they are simply the prophylls of shoots that bear flowers. Nevertheless, in many monocots that have single adaxial prophylls on vegetative shoots the bracteoles are depicted as being lateral, or paired, or even abaxial (Eichler 1875; see also Bruhl 1995; Buys & Hilger 2003); prophyll number and position in monocots will clearly repay further study. For genes controlling floral prophyll development, see Masiero et al. (2004). Bracts and bracteoles cannot be recognised when the flower is single and terminal.

Pedicel articulation may be a useful character, especially in monocots (for information there, see Schittler 1953; Kubitzki 1998a); there is a line across the pedicel where non-fertilised flowers will abscise. The part of the pedicel above the point of articulation may be considered different from the part below, perhaps being more strictly part of the flower proper; the term "pericladium" has been used.

Flower.

The description of the flower proceeds from the outside in, from the sepals or perianth to the carpels, ending up with mention of the stigmatic surface. The only general positional terms used to describe the positions of the parts of the flower are the ad/abaxial and median/lateral pairs. Particularly valuable sources of information for basic floral morphology are Eichler (1875-1878), Engler's Die natürliche Pflanzenfamilien, and the great Flora brasiliensis (Martius & others 1840-1906). Batygina (2002) has edited the first of three volumes on terms used to describe flowers and floral embryology, and there are a number of summaries of aspects of floral morphology and development, including those of Endress (2005a [especially development], 2005c). I have not grappled with the complexities of floral, particularly carpel, vascularization, although the vascular supply to individual organs is frequently mentioned.

There are a few structures that look to all intents and purposes like flowers, even on close inspection, but in fact seem to have evolved by the aggregation of several highly reduced flowers - a good example is Cercidiphyllum. Such extremely reduced floral structures are called pseudanthia (e.g. Claßen-Bockhoff 1990, 1991: Endress 1994b). However, there are all intermediates between this condition and cases like species of Viburnum, with enlarged flowers on the margin of a flat-topped inflorescence all flowers which open simultaneously, and the capitulum common in Asteraceae made up of numerous small flowers which similarly may be functionally a single unit when it comes to atracting the pollinator. In both these latter cases individual flowers are perfectly easily recognisable (see especially Claßen-Bockhoff 1990 for images). There has been a recent resurgence of interest in pseudanthia, focusing particularly on taxa like Lacandonia (Triuridaceae) and on members of Alismatales (e.g., Rudall 2003b; Buzgo et al. 2006 for literature), and the definition of pseudanthia has sometimes been broadened to include strucutres that are part flower, part inflorescence (Rudall & Bateman 2003; Sokoloff et al. 2006). These latter, which may be terminal (Sokoloff et al. 2006), can perhaps be confused with peloric flowers (Buzgo et al. 2006: see below).

Floral development. Many aspects of floral development, whether at the genetic or gross morphological level, are integrated into individual characterisations and throughout the hierarchy and are also discussed separately below. In terms of general patterns, development follows the sequence in which the parts are borne in the flowewr (from outside in) sepals - petals - androecium - gynoecium, with the primordia, at least those of different whorls, all being separate. However, within the androecium is particular there is substantial variation, e.g. centripetal (oldest stamens outside) versus centrifugal (oldest stamens inside: e.g. Corner 1946b, see below) and diplostemonous (sepalline whorl outside the petalline whorl) versus obdiplostemonous (vice versa) development. In a few cases the androecium is initiated before the corolla (Bello et al. 2004 for references) or the two are initiated as common primordia. Finally, Lacandonia (Triuridaceae) is the only angiosperm in which the carpels surround the stamens (see Rudall 2003 and Ambrose et al. 2005 for literature: if the flower is really a pseudanthium the exception goes away, but something like heterotopy is a more likely explanation).

Buzgo et al. (2004) in their study of Amborella note that distinctions between different kinds of floral parts can be hard to make, even the distinction between perianth and prophylls. As Endress (2005c) emphasises when discussing petals, there are a number of features - position, function, development, shape, anatomy, histology, gene activity, and relationships to other taxa that clearly have petals - that "make" a petal a petal, so if a structure does not have one of these criteria, is it thereby excluded from petalhood? Or if it does have one of these features, is it to be included? Thus Maturen et al. (2005) recently found that floral organ diversity genes (B and C) were expressed in the large, white dogwood (Cornaceae, Cornales) bracts, and Geuten et al. (2006) suggest that heterotopic SEP3-like gene expression - in bracteoles and calyx in extant members - was present in the common ancestor of Balsaminaceae et al. (Ericales) clade.... Complications also arise in cases where the petals are believed to be modified stamens (Caryophyllaceae), whether or not these modified stamens also represent nectaries (Ranunculaceae); such complications are discussed where they occur, although it may be noted here that in some classical theories of floral evolution it was thought that the petals of eudicots at least, if not broad-leaved angiosperms as a whole, represented sterilised and modified stamens. Ideas that some flowers may really be pseudanthia (see above), or that evolution of floral parts has been mediated by homeotic changes (e.g. Ronse De Craene 2003; Rudall 2003), further confound our attempts to peg specific terms to the parts of the flower. The general plasticity of perianth parts compounds the general difficulty of mapping of perianth evolution (for example) on the tree (e.g., Endress 2004; D. Soltis et al. 2005b). Finally, there is a kind of "pull of the present" in operation. We commonly talk about the evolution of petals because core eudicots, which are so diverse and on which so much work is being carried out, have them, but what has evolved is more a sharp distinction between organs that function as sepals alone and those that function as petals alone from organs in which these functions were not so shrply distinct.

The drawings of floral development made by Payer (1857) can still be consulted with profit.

Cortical vascular bundles in the flower are mentioned only when they occur (for information, see Ronse Decraene 1992).

Flower part insertion. Flower parts are most commonly whorled, but they may be spiral or even chaotically arranged (Ronse Decraene & Smets 1998a). The distinction between a whorled and a spirally initiated androecium is not always sharp (see Ronse De Craene et al. 2003 for a survey of variation in this character in "basal" angiosperms). Even in a multistaminate, spiral androecium, the outermost stamens may be initiated in pairs, at least when it comes to timing of initiation (e.g. Ronse Decraene & Smets 1993).

In many monocots, although much less frequently in the commelinids than elsewhere in the clade, each of the six stamens is borne opposite a tepal, and the bases of the two tepal whorls do not each separately completely surround the flower; such flowers are as it were pseudo-uniseriate when it comes to the perianth (see also below). Here, unlike the situation in broad-leaved angiosperms, a stamen may be strongly adnate to each tepal of both whorls, both when each tepal is more or less free (Androcymbium [Colchicaceae] is a particularly striking example) and even when the tepals form a tube. Posluszny et al. (2000) discuss a perianth-androecium module in the context of the floral evolution of Alismatales (see also Buzgo 2001; cf. Rudall 2003), and I have (rather uncritically) applied a similar idea to non-commelinid monocots in their entirety. Endress (1995b, Fig. 6) describes petals and stamens arising from a common primordium in Zingiberaceae, and this also occurs in some groups of core eudicots, e.g. Myrsinaceae and their relatives (see Ronse Decraene et al. for a summary; de Laet et al. 1995 for discussion). However, in these cases there is a clearly completely biseriate perianth, a different situation from the condition plesiomorphic in monocots. A further complication occurs in Acorales, some Alismatales and perhaps Piperales. There the bract and the median tepal of the outer whorl are similar developmentally, the median (abaxial) tepal of the outer whorl being bract-like (Buzgo & Endress 2000; Buzgo 2001).

Meristicity. Especially when the parts of the flower are whorled, members of the different whorls, particularly the calyx and corolla, commonly also the androecium, and rather less frequently the gynoecium, have the same basic number; 2- or 3-carpellate gynoecia in otherwise 4- or 5-merous flowers are common in core eudicots. Meristicity of basal angiosperm clades and the eudicots (but less in core eudicots) is notably labile, and a trimerous perianth - and less fequently androecium - is widely scattered (e.g. Kubitzki 1987; Endress 1987; Drinnan et al. 1994; Albert et al. 1998; Soltis et al. 2003; Zanis et al. 2003; P. Soltis et al. 2004; D. Soltis et al. 2005). Monocots are overwhelmingly trimerous in all five floral whorls, unlike the situation in other angiosperms (some Aristolochiaceae are cloase), even if the simplistic distinction (which I used to teach) of monocots - overwhelmingly 3-merous flowers; dicots - often 5 merous flowers, needs qualification. In many core eudicots there is inter- and even infraspecific variation in meristicity although a particular family may be predominantly either 4- or 5-merous.

Floral Orientation. The main variation in the orientation of the flower is whether the median ("odd") sepal, or the median member of the perianth or outer tepalline whorl, is adaxial or abaxial. The plesiomorphic condition for the angiosperms seems to be to have the median member adaxial, although ascertaining this is complicated by the haplomorphy of the flowers of some of the basal pectinations. In the monocots the median member is abaxial (this is associated with variation in prophyll and bracteole position, q.v., i.e. if this/these are absent, the median tepal may switch positions 180 degrees; see also Hofmeister's Rule), but Spichiger et al. (2005; see also Eichler 1875-1978) suggest that Agave, Tillandsia, Allium, Galanthus, Dioscorea, and Lilium all have the median member adaxial. Indeed, exceptions to the rule in monocots are often systematically and biologically interesting, often - but by no means always - being associated with various forms of monosymmetry (see below). Stuetzel and Marx (2005) note the variability of the position of monocot bracteoles/prophylls, perhaps because axillary flowers may in fact represent reduced racemes. For additional discussion of tepal orientation in monocots, see also Eichler (1875), Engler (1988) and Remizova et al. (2006b); the orientation of the flower may depends on the presence of a prophyll, also on the existence of other foliar structures on the pedicel.

Inverted flowers may arise in three ways. 1. The distinctive orientation of the flower may be evident from the beginning of development (Bremer et al. 2001 describe variation in the corolla orientation in monosymmetric flowers of euasterids that includes this). 2. Inversion may be secondary, by resupination. Here the pedicel and/or ovary (when inferior) may twist and become resupinate (as in Orchidaceae), or the corolla itself may twist (Daniel & McDade 2005), or the flower may basically just flop over. This latter condition occurs in Balsaminaceae, for example, which has rather slender pedicels and strongly monosymmetric flowers where most of the mass of the flower (in the form of a large, spurred, median sepal) is initially on the adaxial side, but in the open flower the nectar spur is held in the abaxial position, the flower having flopped over. 3. The individual flowers may be normally oriented with respect to the axis that bears them, but because of how the inflorescence itself is held, they are presented inverted. Monosymmetric flowers that have become inverted by one of these mechanisms seem to be notably common in monocots.

Symmetry. Flowers may be monosymmetric (= bilateral, zygomorphic), disymmetric (two main planes of symmetry), haplomorphic or polysymmetric (= radially symmetrical, actinomorphic - see Neal et al. 1998; Endress 1999). The last two conditions are often lumped together; haplomorphy refers to situations where the flowers appear to be polysymmetric, but the two halves are not mirror images because the parts of the flowers are numerous and spirally inserted. The plane of symmetry in monosymmetric flowers is usually vertical, passing through the middle of the adaxial sepal and of the abaxial petal (or vice versa, as in some monocots), oblique, passing between the adaxial and adaxial-lateral sepals and the abaxial and abaxial-lateral petals, strongly oblique, passing through the middle of an adaxial-lateral sepal and of an abaxial-lateral petal, or, rarely, transverse, with the plane being at right angles to the median plane of the flower. The direction of the plane of monosymmetry in obliquely monosymmetric flowers may vary, even in quite closely related taxa, for instance, whether monosymmetry runs from the odd petal to the odd sepal or vice versa (Ronse Decraene et al. 2002). Flowers with oblique (s.l.) or tranverse symmetry are quite common (cf. Neal et al. 1998). The term monosymmetry thus covers a multitude of sins, especially when details of the lobing patterns and the orientation of the flower (for the latter, see above) are thrown in (cf. Donoghue et al. 1998; Ree & Donoghue 1999). As to lobing in the asterid 1 and 2 groups, although a 2:3 lobing (two corolla lobes make up the adaxial part of the corolla, three make up the abaxial lip) is common in Lamiales in particular (e.g. Weberling 1989), there are many other types of lobing, including 0:5 (e.g. Cichorium. A particularly distinctive type of monosymmetric flower is where the corolla appears to be split down one side, split-monosymmetry, as in a number of Asterales, but also sporadic elsewhere, as in Haemodoraceae, Loranthaceae, etc.; note that the monosymmetric flower of Anigozanthus (Haemodoraceae) and other plants where the whole perianth is involved (Proteaceae) must necessarily be slightly oblique, this is not true when the corolla alone is involved, as in Asteraceae, Loranthaceae and Lamiaceae. For discussion on the evolution of lobing of monosymmetric flowers, see e.g. Endress (2001); for suggestions that monosymmetric flowers in different clades of the asterid I group arose in different ways, see Reeves and Olmstead (2003). The monosymmetry in flowers of Melastomataceae and Lecythidaceae barely involves the corolla at all, just the androecium. For the distribution of enantiostyly, the deflection of the style to one side of the floral axis, see Jesson and Barrett (2003). Asymmetric flowers are uncommon, although they characterise Cannaceae and Marantaceae; in the latter family, at least, the asymmetric flowers may form mirror image pairs, effectively a strong form of enantiostyly. There may be repositioning of some of the parts of the flower after it opens making it more or less monosymmetric. An extreme example is Passiflora mucronata (Sazima & Sazima 1978); such flowers are not considered to be monosymmetric here (cf. Endress & Matthews 2006a). Particularly valuable sources of information for this whole suite of characters are Eichler (1875-1878) and the Flora brasiliensis (Martius & others 1840-1906).

Rudall and Bateman (2004) survey the evolution of monosymmetry in monocots, where recent work has emphasized that taxa like Acorus, Triglochin, etc. with small and apparently undistinguished flowers, may be monosymmetric (Buzgo et al. 2006). Furthermore, detailed studies of development often suggest symmetries other than those evident in the adult flower, in particular, slight monosymmetry may be evident early in development, even if not in the flower at anthesis (e.g. Ronse de Craene 2005). The significance of such findings, from both the point of view of zygomorphy and its involvement in specific pollination mechanisms and the very evolution of (and conceptualization) of zygomorphy is hard to evaluate.

Associated with monosymmetry is tactile/color patterning of the corolla/perianth. This is commonly centered on the median petal, as in the median abaxial petal of many asterid I taxa (e.g. Lamiaceae, Plantaginaceae, scattered in monocots). This is perhaps connected with the orientation of the incoming pollinator and the importance of the lower part of the visual field for shape recognition when the colours have long wavlengths (see Neal et al. 1998). However, in asterids, patterning is centered on the two adaxial petals of the normally-oriented flower of Collinsia (Plantaginaceae), which has a very distinctive pollination mechanism for Plantaginaceae (it is a papilionoid flower). Patterning also occurs on the adaxial petals alone of some Pelargonium (Geraniaceae), the adaxial petal in several Pontederiaceae, Rhododendron (Ericaceae: flower inverted), Schizanthus (Solanaceae: flower inverted), the two adaxial inner tepals alone in Alstroemeria (Alstromeriaceae: flowers presented inverted), and the adaxial petal in some Fabaceae-Faboideae (flower inverted). This may be connected to the fact that, as Neal et al. (1998) note, in ultraviolet light the upper part of the insect's visual field is most important in the recognition of shape.

Pollination syndromes are mentioned only occasionally below (for a summary, see Faegri & van der Pijl 1979; cf. also Waser et al. 1996). In general, similar syndromes have evolved many times, indeed, in the past over-reliance on similarity in pollination syndromes as an indicator of relationships has caused problems from the ordinal level down. Variation in pollination mechanisms is of course extensive, even if characterization of the syndromes of characters that are supposed to characterise e.g. ornithophilous flowers may be overly simplistic (Waser & Ollerton 2006, for refernces). Comments on particularly distinctive or pervasive pollination mechanisms are sometimes made after the family characterizations; flowers that produce oils as a reward are relatively uncommon and tend to be rather restricted in their taxonomic distribution (e.g. Vogel 1974, 1986, 1990). Interestingly, oil sources within the same flower may be in two separate locations (Vogel 1984) if the oil is taken up by the legs of the pollinator (see also Pauw 2006 for similar morphologies within Orchidaceae - Coryciinae). Seymour (2001) lists taxa in which thermogenic flowers (i.e., with respiratory heat production) have been found. Deceit pollination - no particular syndrome, of course - is perhaps surprisingly common (Renner 2006).

The hypanthium is defined here as a disc- or cup-like structure more or less surrounding or borne on top of the gynoecium, but separate from it, and bearing perianth and stamens on its margin. One can think of it as representing receptacular tissue. If the hypanthium has become adnate to the gynoecium, the latter is simply described as being inferior; I never describe the hypanthium as being adnate to the ovary (see also below). Many monocots have stamens adnate to a tube formed by adnation of the two whorls of the biseriate perianth, the perianth tube (see below); I have not used the term "hypanthium" for this condition because the tube is strictly appendicular in origin. However, Velloziaceae and Hydrocharitaceae may have structures more like a "true" hypanthium.

A perianth. The term perianth is often used to refer to all the more or less foliaceous structures enveloping the flower. A perianth in the strict sense (P in the characterisations) occurs when there is only a single whorl of foliaceous structures surrounding the flower, or the parts are all similar and in a single spiral, actually a rather different condition from the first. In a bi- or multi-whorled perianth, all parts may be similar and called tepals - T, or very frequently perianth differentiation occurs, and a calyx (K, made up of sepals) and corolla (C, made up of petals) are distinguishable; differentiation of the perianth has occured many times (see below under petals). Note that the sepals and petals found in some monocot flowers and those of a core eudicot flower are equivalent only by designation, as are those in e.g. Saruma (Aristolochiaceae) or Papaveraceae and core eudicots; there has been extensive parallel differentiation of sepal/petal function. Many monocots have biseriate petalline tepals, and in Lilium B-class floral homeotic genes are expressed in the outer whorl alone (Kanno et al. 1999); in eudicots such genes are expressed only in the inner (petalline) whorl. However, Ambrose et al. (2000) suggest that flowers in Poaceae can be directly compared with those of core eudicots, with the palea, at least, being calycine and the lodicules corolline. This argument tends to emphasise gene expression, so favoring Remane's criterion of special properties (Remane 1952) over other criteria such as position; in such situations, questions as to what the parts "are" become close to unanswerable. The evolutionary and developmental relationships between the arrangement and differentiation of the parts of monocot and eudicot flowers remains unclear.

An epicalyx is of sporadic occurence, and is mentioned only when it occurs. It origin is various; it may represent modified stipules of the calyx lobes (as perhaps in Rosaceae - Rosoideae) or inflorescence bracts (Malvaceae), etc.

Numbers of stamens, tepals, sepals or petals are mentioned only when they are other than what might be expected from the basic meristicity of the flower.

Calyx and corolla aestivation is mentioned only when it is other than cochleate in the broad sense. Thus if there is no mention of valvate or contorted aestivation, or of flowers with open development, it can be assumed that aestivation in the group is predominantly some version of cochleate - and in many families with predominantly cochleate aestivation other types also occur. (Note that imbricate covers all aestivation types in which the edges of adjacent perianth members overlap one another [cochleate here], and also, for example, buds in which the bud scales are overlapping). Endress (1999) observes that in asterid families with contorted aestivation, the direction of contortion is usually constant (an exception is Apocynaceae), whereas in other families it is usually variable. In Lamiales and Fabaceae variation in whether the adaxial or abaxial petals(s) of cochleate corollas overlap the others (descending or ascending cochleate) affords phylogenetically useful distinctions. In core eudicots apart from asterids and Santalales the sepals, whatever their aestivation, usually enclose and entirely cover the petals until, or just before, the flower opens, but in the two groups mentioned the corolla is evident in the bud stage well before the flower opens (Matthews & Endress 2002).

Connation. It can be assumed that the calyx and corolla are free unless otherwise mentioned. Erbar and Leins in particular (Erbar 1991; Leins & Erbar 1997; Leins 2000; Leins & Erbar 2003b) have emphasized the different ways in which a corolla may become sympetalous - early and late corolla tube formation are the two extreme conditions, but intermediates ("early/late") are known. In late corolla tube formation the corolla tube is formed by the confluence of originally free petal primordia, while in early corolla tube formation the part of the corolla with stamens adnate to it is evident as a ring primordium from early on. In the former petals may lag behind stamen development, while in the latter they grow rapidly and soon cover the stamen primordia (Ronse Decraene et al. 2000). A tubular structure is sometimes formed by both filaments and corolla lobes all forming integral parts of the tube, as in some Convolvulaceae and Diapensiaceae. Calyx and corolla together may form a tube, but this is rare in eudicots (e.g. Passifloraceae); a rather similar condition where the tube is formed from the tepals of both whorls is quite common in monocots (see above). Indeed, in monocots as a whole there is considerable variation, even between closely related taxa, both in the degree of connation of the tepals and of the adnation of the stamens to the tepals (see also Weberling 1989).

Perianth/sepal/petal/tepal Venation. Sepals in core eudicots are often innervated via a three trace:three gap vascular supply, or the lateral sepal traces come from the petal traces, etc. However, if the sepals are much reduced, there may be either a single trace, or no trace at all, as in some Ericaceae-Rhododendroideae and Oleaceae. In the reduced flowers of Morus (Moraceae), tepals of staminate flowers may have a single trace, those of carpellate flowers, three traces (Bechtel 1921). The sepals of other eudicots and of members of the ANITA grade sometimes have only a single trace (e.g. Glück 1919; von Gumppenberg 1929; Kaussmann 1941; Hiepko 1964; von Balthazar & Endress 2002a). Petals often have a single trace, although those in e.g. Campanulaceae represent fused branches from adjacent sepal bundles, while the single traces in many Asteraceae are in fact commissural bundles. The single trace often shows a complex branching pattern. Members of both tepal whorls in monocots commonly have three traces, but sometimes each member has only a single trace - perhaps particularly frequently in Asparagales (Gatin 1920; Kaussmann 1941; Utech e.g. 1984, 1987, and references) and Alismatales. Since I am unclear over both the pattern and hence perhaps of the significance of the variation in this character, I have put in more detail than may ultimately be necessary.

Petals. Only distinctive features such as whether or not the petals are clawed, i.e. with a sharply narrowed and more or less elongated basal portion, or ligulate, with more or less petaloid productions on the adaxial surface, are mentioned. From the above remarks, it is very probable that sharply differentiated petals have evolved several times, even if they show similar patterns of gene activity (as in Aristolochia and Arabidopsis - Jaramillo & Kramer 2004; see also Kramer & Irish 1999; Zanis et al. 2003; etc.). Monocots are described as having petals only when the perianth is differentiated into two morphologically distinct whorls, the outer being more or less green, smaller, and protective. The terms corolla or petals are not used to refer to the undifferentiated perianth members of the large flowers of taxa like Agave, Hemerocallis, or Blandfordia, brightly colored although they may be; these and many other monocots have petaloid tepals.

Nectaries. Smets (1986, esp. 1988) provides a useful general survey (see also Vogel 1977; Weberling 1989) of floral nectaries. In core eudicots nectar may be secreted from a more or less annular, disc-like structure with stomata through which the nectar emerges. This commonly immediately surrounds the ovary, or it is on top of it, if the ovary is inferior. Although "discs" and nectaries are frequently noted in the eudicot characterisations it is by no means certain that all have a similar anatomy, etc. (e.g. see Smets 1986; Vogel 1997). Smets et al. (2003) distinguish between receptacular nectaries, characteristic of rosids, and gynoecial nectaries, typical of asterids (the name refers to where the nectaries are found, see Smets 1988 for a distribution of the main nectary types on a Dahlgrenogram). However, note that the CRABS CLAW gene is expressed in both floral and extrafloral nectaries in the few core eudicots studied, and would appear not to pay particular attention to this distinction (Lee et al. 2004). Nectar may also be secreted by hairs on the calyx or corolla or more generally from folds or other tissue on the tepals or petals, from what are clearly modified stamens or parts of stamens, and even from the surface of the gynoecium. Brown (1938) provides an early but fairly comprehensive survey of nectaries; whether or not features like nectary vascularisation (as in e.g. Capparaceae, Passifloraceae, Salicaceae, and Cucurbitaceae) and/or the presence of stomata in the epidermis covering the nectary have systematic significance is unclear. Smets et al. (2000) survey monocot nectaries in general, and Rudall (2002) and Remizowa et al. (2006a) discuss the diversity of septal nectaries, epithelial nectaries in the septal position that occur in many monocots (see also Daumann 1970; van Heel 1988). Septal nectaries seem not to occur when placentation is parietal (Rudall et al. 2005) and of course when the gynoecium is apocarpous (it is difficult to have septal nectaries when there are no septae...); there is a correlation between septal nectaries and the occurence of some postgenital fusion of the gynoecium (e.g. van Heel 1988; Remizowa et al. 2006a). (Some broad-leaved angiosperms have cavities in the ovary septae, but these are not known to secrete nectar [Ronse Decraene et al. 2000b]). Nectaries are quite frequently lost, and not only in connection with the development of the wind pollination syndrome. The presence of a nectary is noted in the appropriate place in the descriptive sequence of the flower, i.e. proceeding from the oustide in; complete absence of a nectary is noted immediately before the description of the gynoecium.

Stamen number. The number given refers to the number of fertile stamens; "many" means that there is a variable and large (more than 15) number of stamens.

Stamen development and position. When there are many stamens, the sequence of initiation may be centripetal (especially the ANITA grade, magnoliids), or it may be reversed, being centrifugal, as is common in taxa with fasciculate androecia. In some taxa that have secondarily numerous stamens the distinction between the two modes of initiation may be unclear (Hufford 1990). A number of taxa with secondarily numerous stamens initially have only five or ten primordia, in the former case, the primordia are often antipetalline (see stamen insertion, below) and the stamens themselves may be more or less connate or in fascicles (see esp. Corner 1946b; also Hirmer 1917; Pauzé & Sattler 1978; Stebbins 1974; Weberling 1989 and references). When there are two staminal whorls, the outer (antesepalous) is normally initiated before the inner (antepetalous), the condition of diplostemony. However, in obdiplostemony (see Eckert 1966) the stamens of the antepetalline whorl appear to be outside the stamens of the antesepalline whorl. Diplostemony is by far the most common, and only obdiplostemony is noted specifically in the descriptions. (I also note an associated feature, the orientation of the carpels [see below].) If the stamens are equal in number only to the sepals or outer perianth whorl, they may be opposite (haplostemony) or alternate with them, in the latter case being opposite to the inner whorl of the perianth or petals (obhaplostemonous: Ronse Decraene and Smets [1995] suggest other terms should be used to describe such features in monocots). Stamens within a single flower may be of different lengths, and this is commonly seen when comparing members of the two different whorls, although it is often only slight. It is more pronounced when the flowers are heterostylous, as in tristylous Oxalis (there is also variation between individuals, as there is in distyly). More importantly from our point of view are situations like those in Brassicaceae, which are tetradynamous, with four long and two short stamens, while Lamiales with only four stamens are commonly didynamous, with two long and two short stamens. The degree of connation of the filaments and/or anthers also varies considerably. Finally, stamens in some Piperales and Alismatales (for example) may be initiated in pairs associated with individual perianth members. Ronse Decraene and Smets (1996a) distinguished between two kinds of stamen pairs, one associated with the process of reduction in number of stamens, the other with an increase in number (dédoublement). By their criteria, Caryophyllales vary in this feature, as do Rosids as a whole.

Stamen insertion. Stamens are usually adnate to the corolla only when the latter is sympetalous, but not all sympetalous taxa have stamens adnate to the corolla (e.g. Ericaceae and a number of other Ericales), and some monocots with free tepals have stamens adnate to those tepals. For insertion of the stamens relative to the gynoecium, see Ovary Position below.

Stamen morphology is very variable, and only some of this variation is mentioned below. Filaments are usually free and slender, but they are sometimes connate, and rather stout filaments is the plesiomorphic condition for angiosperms. There is usually a single vascular bundle, but there are three in many Magnoliales. The relationship between the sporogenous tissues of the anther and the supporting tissue of the filament provides a further useful suite of characters. The stamens may be laminar, with the anthers more or less embedded in one surface (as in many magnoliids), or they may be stout, with little distinction between the anther with its well-developed connective and the filament (in most Ranunculales, see also Rousseaceae, etc.), or the filament may be quite narrow and be attached to the base of the anther, i.e., it is basifixed. The actual point of connection may be further more or less narrowed (Nothofagaceae), or hardly narrowed, although the base of the anther thecae may still be quite distinct from the filament (e.g. Pittosporaceae, Escalloniaceae, etc.), or the filament may join the anther in what appears to be a basal pit (esp. in Saxifragales). In general in core eudicots, the anther/filament junction is often more or less dorsal (dorsifixed). Finally, Rudall (2001b) summarizes the distribution of centrifixed anthers in which the slender filament tip is inserted into a hollow in the anther in monocots (cf. Saxifragales above). It can be very difficult to distinguish clearly between these various "types" of stamens, and my treatment of this set of features leaves something to be desired. The connective may also be distinctly prolonged, often a systematically valuable feature. Hufford and Endress (1989) and Endress and Stumpf (1991) provide invaluable recent surveys of stamen morphology; Wilson (1942), Dahlgren et al. (1985), Endress (1994b), and others summarise variation.

Walker-Larsen and Harder (2000) provide a review of staminodes and their phylogeny. Staminodes vary morphologically from stamens that appear functional but produce non-functional pollen to minute and strictly rudimentary structures as well as complex petalline (especially in some Zingiberales) or nectar-secreting (?Ranunculaceae) structures. The degree of development of staminodes, whether in perfect flowers (e.g. Lamiales, see Bignoniaceae, Plantaginaceae, Calceolariaceae, etc.: Endress 1998) or in carpellate flowers, can vary considerably within a family, although others consistently never have staminodes.

Anthers are usually tetrasporangiate, with two parallel thecae each with two sporangia, but they are sometimes bisporangiate, in which case it is the two sporangia of a single theca that have usually been lost, the stamens then being unithecate (see Weberling 1989; Endress & Stumpf 1990). Anthers vary for example in length (cf. Achariaceae and Salicaceae) and whether or not the thecae (= sporangium pairs) are laterally or apically confluent, or are superposed, or are in part sterilised, etc. (see Trapp 1956a, b, for Lamiales in particular). Anther dehiscence is an important character. Anthers are usually introrse, opening internally, but in a number of taxa they are extrorse (there may be variation in this in the different anther whorls of a single flower in Lauraceae!). Each theca normally opens separately by longitudinal slits, less frequently by slits common to the two thecae, the thecae then being apically confluent, or by flaps (valvate) or pores (porose); only the uncommon conditions are mentioned. Porose anthers, or pollen coming out the top of a cone formed by all the anthers, are common in buzz-pollinated plants; flowers of species with the buzz pollination syndrome can look very different from those of their close relatives with different pollination syndromes and quite frequently have also lost their endothecium (see below, Anther Wall).

Secondary pollen presentation. In a number of taxa pollen is picked up by the pollinator from petals or styles, or as it is forced out of a tube formed by connate anthers by the elongation of the style (Yeo 1993; Leins 2000), that is, pollen is presented to the pollinator other than on the individual anthers. Although such secondary pollen presentation tends to characterise groups of related genera, it has evolved many times and in a number of ways (perhaps most notably in Asterales: see Leins & Erbar 2003b). Indeed, one of the most spectacular examples of secondary pollen presentation I have seen was in a species of Dombeya (Malvaceae), a genus in which I never expected to see such a pollination mechanism; the pollen was attached to the staminodes.

A number of families, especially in Lamiales and Solanales, have placentoids, a parenchymatous outgrowth of the connective into the anther loculi (Hartl 1964; Endress & Stumpf 1991: also some Caesalpinioideae, Cannaceae, Zingiberaceae, etc.). The detailed distribution of this character is unclear.

Anther wall. The anatomical structure of the anther wall shows much variation, some clearly directly linked to variation in anther dehiscence. Most anthers of angiosperms have a hypodermal layer with distinctive lignified thickenings, the endothecium. There is variation in endothecial development. Both secondary parietal cell layers (products of periclinal divisions of the primary parietal cells, themselves the product of the first division of the archesporial cells) of the anther wall normally divide further ("basic type" of wall development), or the inner cell may remain undivided (the "dicot type"). The endothecium can also develop directly from undivided outer secondary parietal cells (the "monocot type" - Davis 1966). Sampling of this character is poor, the distinction perhaps dangerously typological, and there may be considerable infrafamilial variation (Hermann & Palser 2000; see also Rudall & Furness 1997; Carrizo García 2003). I have not described or analysed variation in the extent of the mature endothecium and endothecium-like tissue; Chatin (1870), Hufford and Endress (1989) and Endress and Stumpf (1991) in particular provide much information on these features. Endress and Stumpf (1991) note a correlation in eudicots, especially in the rosid area, between the development of endothecial-like cells around the connective when the latter is well developed, and around the individual thecae when the connective is less well developed. Individual taxa like Haptanthus show massive development of the endothecium (Doust & Stevens 2005). There is also much of interest in details of the elongation of the endothecial cells relative to the long axis of the anther and in the thickenings of endothecial cells (e.g. Noel 1983; Manning 1996), but these, too, are rarely mentioned below. in porose anthers that have lost their endothecium, epidermal cells are often notably thick walled and form an exothecium; an exothecium is also present in the anthers of many gymnosperms.

Microsporogenesis. Data are taken in part from G. Dahlgren (1991) and Johri et al. (1992). In simultaneous microsporogenesis the microspores initially form tetrahedral tetrads, in successive microsporogenesis (common, but not universal in the monocots - see Furness & Rudall 2000a, 2001b), the microspores are initially linear, but they become tetragonal. The occurence of simultaneous microsporogenesis in monocots is correlated with trichotomosulcate pollen (Rudall et al. 1997), however, Furness et al. (2002b) caution against being overly typological when describing microsporogenesis, Schols et al. (2005) note that aperture morphology and microsporogenesis are not necessarily linked, as was confirmed by Sannier et al. (2006), who also noted infraspecific variation in microsporgenesis. The tapetum is usually glandular/secretory/parietal, with separate cells, and only rarely amoeboid/plasmodial, forming a syncytium, the latter being commonest in monocots (Furness & Rudall 1998, 2001b); for a general survey, see Pacini et al. (1985) and for possible delimitation of states, Pacini (1997). Taxa with a glandular tapetum usually have orbicules or Ubisch bodies, although they are absent from many Rubiaceae and some Annonaceae, Fabaceae, etc. (Keijzer 1987; Huysmans et al. 1998; Furness & Rudall 2001a; Wang et al. 2004). Details of orbicule morphology and development may be of systematic interest (e.g. Vinckier & Smets 2002a). The number of nuclei in tapetal cells varies from 1 to 6 or more, but I have not handled this character well. The common situation seems for the tapetal cells to be binucleate, although some cells may be uninucleate. Wunderlich (1954) summarises information on this character; there is clearly much variation within families, and also differences between observers. Wunderlich also noted extensive variation in the ploidy level of the tapetal cells.

Much attention has been paid to details of pollen grain morphology as indicators of relationships. Here I refer consistently only to the number and nature of the germination pores, or to their absence. Pantocolporate or pantoporate grains have apertures all over the surface. Porate pollen in broad-leaved angiosperms may be associated with wind pollination (Endress & Stumpf 1991). Although most monocot families have mono(ana)sulcate grains, details of the developmental pathways of such grains may vary (Penet et al. 2005), and they often have other types as well, although I rarely mention these in the characterisations unless they are common within a clade. Operculate pollen is scattered (Furness & Rudall 2003). Comparable variation in eudicot groups is also rarely mentioned. Inaperturate pollen grains - in many cases, e.g. Araceae, perhaps more properly referred to as omniaperturate - are scattered, but characterise some monocot families and Zingiberales as a whole (Furness & Rudall 1999, 2000b, 2001b).

Details of pollen microstructure, whether visible under S.E.M. or T.E.M., are mentioned only inconsistently; see Harley and Zavada (2000) for an attempt to think of pollen variation in monocots as a whole in the context of phylogenetic analysis. However, as Walker and Doyle (1975, p. 677) note, "probably no other palynological character has been responsible for so much terminological confusion as pollen wall morphology". Thus although the term ektexine as originally defined (Erdtman 1943) was more or less synonymous with sexine, it is now invariably used in the redefined sense of Fægri (1956; see also Faegri & Iversen 1964). Unlike ektexine, sexine does not include the foot layer. Sexine and nexine are distinguished on purely morphological criteria, whereas ectexine and endexine differ in their staining properties. The two sets of terms are therefore suited for slightly different applications. Pollen grains are usually resistant to acetolysis; groups that are exceptions to this, like many Zingiberales and Laurales, are noted. Endexine variation needs attention. In gymnosperms and at least some angiosperms it is lamellate (e.g. some Annonaceae, Acoraceae) or compact; it is usually absent in monocots, but is spongy in Araceae (M. Weber et al. 1999).

Data on the number of nuclei in the pollen grains at the time of their dispersal are taken mostly from Brewbaker (1967). The basic condition is to have only two nuclei (associated with a sporophytic incompatability system), however, there may be three nuclei, of which two are gametes (this is associated with gametophytic incompatibility systems - see Wheeler et al. 2001). Zona (2001) surveyed the distribution of pollen grains containing starch in monocots, and Baker and Baker (1979) and Franchi et al. (1996) give further details for angiosperms as a whole. There are suggestive distribution patterns of taxa that commonly have starch, although sampling needs to be improved; it is fairly common for the odd taxon in an otherwise starch-free family to contain starch, and vice versa.

Carpel number. This is usually quite easy to ascertain, with the number of lobes of the stigms or style often being useful confirmatory evidence. However, in Primulaceae, etc., with free central placentation the stigma may give little indication of the carpel number. In a number of taxa apparently with but a single carpel, it is unclear whether there is really only a single carpel or the gynoecium is reduced from a syncarpous condition, i.e. it is pseudomonomerous (for discussion about monomerous gynoecia, see e.g. Eckardt 1937). Carpel number in Papaveraceae and Brassicaceae and their relatives and in Berberidaceae has long been a bone of contention, and Brückner (2000, p. 273) expressed "her serious hope that she may be the last to have invested time and effort in refuting the 2n and 3n theories [basically, that the gynoecium is made up of two or three times the number of carpels it appears to be], at least for Berberidaceae, Papaveraceae, and Capparales." "Many" in the characterisations means that there is a variable and large (more than 15) number of carpels.

Carpel development and closure. Carpel primordia are initially U-shaped, and if the area between the arms of the U becomes meristematic, the meristematic cross-zone, the carpel develops as if it were a tube, the ascidiate condition. If no meristematic cross-zone develops, the carpels are plicate. Intermediates exist, and even fully ascidiate carpels may have elongated stigmas, which one might think were the signature of the plicate carpel condition. Furthermore, ovule position is not tightly correlated with carpel type (Endress 2005b). Particularly in angiosperms of the basal pectinations, rarely elsewhere, the carpel margins are initially open (and ascidiate), but post genital occlusion may occur directly by tissue fusion or indirectly by secretion of the cells of adjacent margins (Endress & Igersheim 1997, 2000 [the latter a particularly useful summary]; Igersheim & Endress 1997; Doyle & Endress 2000); for this character optimised on trees, see e.g. P. Soltis et al. (2004), D. Soltis et al. (2005b). Taylor (1991) discusses this and many other aspects of carpel morphology. Although angiosperms are characterised by their closed carpels, a number of taxa have young carpels that are open and have exposed developing ovules (Tucker & Kantz 2001 for a summary). Furthermore, in a few taxa the carpels open as the seed develops, so the plant then appears gymnospermous. This variation is sporadic, and such carpels are found in Dioncophyllaceae, some Berberidaceae, Malvaceae-Sterculioideae, etc.

Carpel vasculature. Carpels normally are supplied by three vascular bundles, but is a few families there are five (Sterling 1969; Dickison 1971 for references).

Carpel connation, syncarpy, is indicated by square brackets placed around carpel number. Details of the degree of syncarpy are often systematically useful, although syncarpy is highly homoplasious (Armbruster et al. 2002). In taxa like Boraginaceae, Ochnaceae and Lamiaceae with a lobed gynoecium and gynobasic style, it can seem that the carpels are free, however, there is a single central style.

Ovary position. When the ovary, i.e., that part of the gynoecium bearing ovules, is inserted on the receptacle above the points of insertion of the sepals, petals, and stamens or the hypanthium (see above), it is superior, this being indicated by a line under the carpel number, and it is inferior when these structures are inserted on top of the ovary. There are intermediate conditions (semi-inferior). The ovary may become inferior by early vertical growth beneath the perianth members making the whole apical meristem concave; initiation of the youngest floral primordia occurs on this concave meristem (appendicular epigyny), or, to put it in a somewhat different way, the inferior ovary forms through congenital adnation of the carpels to the bases of the perianth members and stamens. The vascular traces to the stamens and perianth members diverge in an acropetal sequence (e.g. Costello & Motley 2004). Alternatively, vertical growth of the periphery of the receptacle may occur after initiation of the primordia of at least sepals, petals and stamens on a normal convex apex, the vascular supply to the appendages is displaced and there are descending traces to the ovules; this is the rather rarer case of receptacular epigyny (e.g. Kaplan 1967; Soltis et al. 2003b). Unless mentioned otherwise, ovaries become inferior by the action of the former developmental pathway. However, note that ovary development of relatively few plants have been studied in detail and whether or not the distinction between the two is really that sharp is questionable (e.g. Dengler 1972; Smyth 2005). Ovaries that are more or less immersed in nectariferous tissue alone, e.g. some Celastraceae, are described here as being superior. I do not use the terms hypogyny, epigyny and perigyny; these properly refer to floral architecture as a whole. Although the change from superior to inferior used to be one of the major evolutionary "trends", we are realising that there is a growing number of cases where the ovary may be secondarily superior. Rhoipteleaceae (Fagales), Opiliaceae (Santalales, but see the discussion there), also Apiales, Asparagales, Asterales, Commelinales, Cucurbitales, Saxifragales and Poales are examples, although most have not been studied in any detail. Saxifragaceae in particular show great variation in ovary position (Soltis & Hufford 2004), while reversion of ovary position from inferior to superior may have occured e.g. in Haemodoraceae (Simpson 1998a, b), Asparagales, etc. There may be an association of such secondarily superior ovaries with parietal placentation, as in Begoniaceae (Cucurbitales), Menyanthaceae (Asterales), and Pittosporaceae (Apiales) (see also Endress 2005a and references).

Carpel orientation. The basic source of data is Eichler (1875-1878: see also Baillon 1866-1892; Le Maout & Decaisne 1868; Goebel 1887; Eckert 1966). When the flower has two carpels, they are usually median, i.e., they are one above the other in the median plane of the flower, but they are sometimes transverse, in the horizontal plane. Only the latter condition is mentioned in the descriptions of 2-carpellate flowers, which can otherwise be assumed to have median carpels. Infrageneric variation in this character, with the carpels being either transverse or median, is known, as in Ribes (Eichler 1878). The arrangement of carpels in 4-carpellate flowers is usually median + transverse, and only the diagonal arrangement is mentioned. When there are the same numbers of carpels as members of the outer perianth whorl, the position (alternate, opposite) of the carpels relative to that whorl is mentioned. Carpels opposite the corolla are common in obdiplostemonous taxa, but they are normally opposite the calyx (or members of the outer whorl of the perianth). When there are three carpels and five sepals, the position (adaxial, abaxial) of the median carpel is mentioned.

Placentation in syncarpous gynoecia is described as being axile, lamellar, parietal, or free central. Basal ovules are nearly always ascending, apical ovules are pendulous. When multiovulate carpels are free, the ovules are often described as being marginal in their insertion. In a few bicarpellate ovaries there is what is called an apical septum. This is supplied by by the dorsal median bundle of the carpel, not the ventral bundle, and is a modification of a gynobasic stylar morphology (see below).

Transseptal bundles are found in the septal radii, and branches vascularise the placentae (Eyde 1988 and references). The normal condition, never mentioned in the characterisations, is for the vascular bundles to be in the placentae themselves. In the placentae of many taxa with parietal placentation there is an inverted bundle immediately underneath a normally orientated (i.e. phloem to the outside) placental bundle (Puri 1946).

Intralocular hairs may occur in the young ovary, these are secretory in monocots (Rudall et al. 1998c) where they are commoner than in broad-leaved angiosperms. In core eudicots the hairs, when they occur, appear to be non-secretory and may be very conspicuous when the fruit is mature, as in Malvaceae.

Ovules. The most useful surveys of aspects of ovule morphology and development are those of Takhtajan (editor: 1985, 1988, 1991, 1992), Kamelina et al. (editors: 1981, 1983), Batygina et al. (editors: 1985, 1987, 1990), Johri et al. (1992), and Danilova (editor: 1996). There are an increasing number of suggestions that ovule terminology needs to be reformed (e.g. Shamrov 1998, see also 2002, 2003, 2004; Endress 2005c), indeed, current terms are in part confusing and arbitrarily delimited. Rudall (1997) surveyed ovule morphology in monocots; Endress and Igersheim (1997, 1999) and Igersheim and Endress (1997, 1998) describe ovule morphology and anatomy in magnoliids and the Anita grade and Taylor (1991) also emphasised these taxa in his survey. However, for a number of taxa, the only information available remains that in early works (e.g. see van Tieghem 1898). Some parasitic or hemiparasitic taxa lack organized ovules, the embryo sac alone being discernable (see Santalales in particular).

Ovule number is always mentioned, although it is often very variable. "Many" means that there is a variable and large (more than 15) number of ovules.

Ovule morphology is important. Ovules can be described as being atropous (= orthotropous), anatropous, amphitropous, campylotropous, circinotropous or hemitropous, depending on the curvature of the body of the ovule and its relation to the funicle. Gifford and Foster (1989) and Bouman and Boesewinkel (1991) emphasize the variety of ways in which an ovule can become campylotropous.

Ovule orientation. The direction of curvature of ovules is usually consistent within taxa where there are few ovules, and ovule orientation is usually mentioned only for such taxa. Epitropous ovules are curved adaxially (= dorsal raphe), apotropous ovules are curved abaxially (= ventral raphe: Björnstad 1970); this variation is independent of whether the ovules are apical or basal. I initially recorded only the apotropous condition, so epitropous ovules are under-reported. Pleurotropous ovules are held sideways so the micropyle faces laterally. When there are many ovules per carpel their orientation is either variable or pleurotropous. Even in families like Sapindaceae and Menispermaceae where there are only two ovules per carpel, ovule orientation can vary within the carpel (e.g. Svensson 1925; Mauritzon 1936). Endress (1994b) suggested that it might be more useful to refer to the curvature of the ovule with respect to that of the curvature of the carpel margins, this curvature will be at right angles to the axis that determines epi- versus apotropy.

Funicle length. Some taxa, most notably in Caryophyllales, have long funicles; these are defined as being about the length of the body of the ovule at anthesis or longer. In other groups, e.g. Sapotaceae and Sapindaceae, the ovule is sessile. When funicle length is not mentioned it can be assumed that it is unremarkable. In taxa with circinotropous ovules the funicles are very long and curled.

Nucellus condition. In crassinucellate ovules the archesporial cell (see below) divides periclinally, whether or not the parietal cell so produced divides again; in tenuinucellate ovules there is no periclinal division of the archesporial cell, which gives rise directly to the embryo sac, and there are no parietal cells between embryo sac and epidermis of the ovule. In a "pseudotenuinucellate" ovule the parietal cell does not divide again (or only once) and also may disappear, leaving the ovule looking as if it were tenuinucellate; weakly crassinucellar (Endress 2003c) is a better name for this condition. A nucellar cap develops when epidermal cells at the apex of the ovule divide periclinally (giving "pseudocrassinucellate" ovules - Davis 1966; also Bouman 1984; Rudall 1997; Endress 2003c); nucellar caps are mentioned only when they occur. Shamrov (2002) discusses nucellus morphology in some detail, while Endress (2003c) not only adds the incompletely tenuinucellar condition in which there are subepidermal nucellar cells laterally to the embryo sac, but not terminally, but also gives extensive information on the distribution of the various kinds of nucellus in Malpighiales and many asterids.

Integument number is an important character; Shamrov (2003) is a useful survey. Ovules usually have one (unitegmic) or two (bitegmic) integuments. A unitegmic ovule may arise either by adnation of these two integuments, or by suppression of one of them (Bouman 1984). The outer integument may be dermal or subdermal in origin, the inner integument epidermal (?always: van Tieghem 1898), and Grootjen and Bouman (1981) and Bouman (1984) suggest that this variation may be of systematic significance, but I have not followed up on their suggestion.

Integument thickness is of systematic significance. Integuments are commonly two or three cells thick over the body of the ovule, and only when they are thicker is this included in the characterisations. The goal in these pages is to record integument thickness as the number of cell layers, both for the mature ovules and also (when it differs) for the maximum number of cell layers produced during the later development of the seed coat. Strictly speaking, multiplicative integuments are those in which the thickness increases after fertilisation (but cf. e.g. Tokuoka & Tobe 2002 who use "multiplicative" as equivalent to "thick" - and this at the ovule stage). The integuments are commonly rather thicker at the micropyle than elsewhere, but this is generally not mentioned. For information on integument thickness, see e.g. Netolitzky (1926), Huber (1969), Grootjen and Bouman (1989), and Johri et al. (1992). In orthotropous ovules the outer integument is often thinner than the inner, and it has been suggested that a thick (outer) integument helps to ensure curvature of the ovules (Endress 2005c). The outer integument at the time it is initiated may be cup- or hood-shaped (Yamada et al. 2001); the significance of this is unclear (but see Endress & Igersheim 2000; Endress 2005c).

The micropyle in the mature ovule, i.e., at the time of fertilisation, shows interesting variation. It may be formed from the outer integument only (exostomal), the inner integument only (endostomal), or both integuments (bistomal), or it may be zig-zag, the integuments (rarely one, in unitegmic taxa) overlapping in such a way that the passage from the nucellus to the outside is not straight: one can think of most ovules with zig-zag micropyles as being variants of the bistomal "type"). Extensive infraspecific variation in micropyle morphology has been observed in Putranjivaceae (Tokuoka & Tobe 1999), and given the sampling of this character, more variation is to be expected. In a few taxa, e.g. some Laurales, the micropyle is not covered by the integuments when the flower is at anthesis.

Embryo sac. The archesporial cell is usually single (exceptions are noted). It ultimately produces a single embryo sac or female gametophyte from the germination of a single megaspore (i.e. it is monosporic in origin, the common condition), itself a meiotic product of a single megasporocyte (see Bouman 1984 for some information). Although the Polygonum-type embryo sac, monosporic and with eight nuclei, the egg cell being adjacent to the micropyle, is commonest in angiosperms and has long been thought to be plesiomorphic, this view is now being contested (see Friedman et al. (2003a, esp. b; Friedman & Williams 2004; Friedman 2006). For an outline of the major variants in embryo sac morphology, see Johri et al. (1992), and for variation in embryo sac morphology and development in basal angiosperms, see Friedman and Williams (2004). Ross and Sumner (2004) suggest that in Arceuthobium americanum the egg apparatus arises at the lower pole of the embryo sac, but they also assume that the ovule is likely to be atropous, rather unlikely in Santalales.

Ovule variants. The nucellus may protrude as a beak through the micropyle (Bouman 1984, for data). Obturators, whether funicular, placental, or integumentary (Shamrov 2004) are closely associated with the micropyle; the ponticulus is a kind of obturator found in some Anacardiaceae. Vascular bundles may occur in the outer or somewhat less frequently in the inner integument; they may be branched or not. They are mentioned in the description of the seed coat. The endothelium (integumental tapetum is a synonym) is a distinctive layer or layers of cells (sometimes up to 32-ploid) of the inner (or only) integument abutting the nucellus; it is commonest in unitegmic taxa (Kapil & Tiwari 1978) and may directly abut the embryo sac due to the breakdown of the intervening cells. The chalaza is variously developed, and there is frequently distinctive vasculature or thickening or staining of groups of cells in the chalazal region (e.g. the hypostase, cf. the epistase). The embryo sac may sometimes enlarge greatly and become haustorial (Mikesell 1990 for a review).

Style position. The style is usually terminal on the ovary and more or less continuous with it. Variants are noted. Thus the style may be distinctly lateral on individual carpels (some Rosaceae-Rosoideae), or the single style of a syncarpous gynoecium may be lateral (Chrysobalanaceae: only one carpel is fully developed, the gynoecium being pseudomonomerous). In some taxa the separate styles of a syncarpous gynoecium may be borne towards the periphery (Hillebrandia - Begoniaceae), not in the center, as is the normal condition in such circumstances; in such a situation the ovary may be described as having a well-developed roof. The style may also be more or less impressed in the apex of the ovary, the extreme condition being a gynobasic style. In such cases, the ovary may secondarily come to surround the style which then appears terminal (Hartl 1962; Jäger-Zürn 2003 and references), however, this condition can be recognised i.a. by the pollen transmitting tissue going down what appears to be the central axis of the ovary to the very base and the distinctive vascularization of the apical septum.

Style connation. I initially tried to make a distinction between styluli, terminal to basal on separate carpels, or the terminal free parts of otherwise connate carpels, and the style, single and borne on connate carpels (e.g. Baumann-Bodenheim 1954). Normally presence of a style would associated with presence of a compitum (see above). Stylodia were branches of a style, be it ever so short. However, the distinction between style, styluli and stylodia in syncarpous gynoecia proved cumbersome and difficult to make, and I now refer simply to styles and branches; however, styluli is a useful term to describe the styles associated with carpels (or carpel apices!) that are manifestly separate. I use these distinctions in the characterisations, although perhaps still not consistently.

Heterostyly is scattered in flowering plants, although particularly common in Oxalidaceae, Primulaceae, Turneraceae, etc. For a summery of recent work on heterostyly, see New Phytol. 171(3). 2006.

Stylar canals. Styles or style branches may be solid or hollow, the central cavity being formed by the schizogenous separation of cells, but this feature is recorded rather erratically. The common condition in eudicots at least is for the style to be solid (Weberling 1989: note that Guéguen 1901 quite often talks about some kind of canals there), however, hollow styles can be difficult to see, because the stylar canal can be very narrow, moreover, the presence/absence of a stylar canal may vary between closely-related taxa (Bensel & Palser 1975a). Capus (1878) and Guéguen (1901) are useful early sources of information; data for asterids are in part taken from Anderberg (1992) and Bhatnagar and Uma (1969). Rudall et al. (2002) suggest that the style in monocots is generally hollow, but again with some exceptions (see also Johri 1966a); for the style in a few monocots, see also Hartl and Severin (1981) and Johri (1966a, b: "Liliaceae").

The stigma shape is variable and sometimes distinctive. It may be punctate, variously expanded, canaliculate or decurrent. Its surface may be wet (quite frequently in taxa with binucleate pollen grains) or dry (grains often trinucleate - see J. Heslop-Harrison 1981; Y. Heslop-Harrison & Shivanna 1977). However, variation here often seems to be between families that are closely related, and the nature of the surface needs further study (e.g. see Igersheim et al. 2001 for conflicting reports in Alismatales). Some taxa have notably papillate stigmas, or the receptive surface may be localised on multicellular papillae or hairs (Y. Heslop-Harrison 1981 for a summary); there is considerable variation within Malvaceae-Malvoideae.

The correlation of stigma surface with incompatability type is not perfect (Y. Heslop-Harrison 1981; Wheeler et al. 2001). Details of incompatability systems are described by Charlesworth et al. (2005). Both gametophytic and sporophytic systems seem to have evolved in parallel, but it has been suggested that the gametophytic system in both rosids and asterids may be identical (e.g. Steinbachs & Holsinger 2002: the stylar response is mediated by a glycoprotein with ribonuclease activity), although at axactly what level it might be a synapomorphy is unclear, largely because so little is known about details of the whole system.

Compitum presence. In many syncarpous gynoecia with at least partly plicate carpels the loculi/placentae are in connection with each other via an intragynoecial compitum, tissue that allows pollen tubes to fertilise any ovule no matter where the pollen grains landed on the stigma, or, if the style in branched, no matter on which stigma they landed (Carr & Carr 1961; Endress 1994; van der Schoot et al. 1995). If the carpels are completely synascidiate there can be no intragynoecial compitum. Some Nymphaeales, Laurales, etc., with free carpels have an extragynoecial compitum, of which the hyperstigma is a variant, and here carpels are placed in contact not by pollen tube transmitting tissue but by secretions of the stigmas through which the pollen grains grow (Endress 1982).

As to pollen tubes, Prósperi and Cocucci (1979) suggest that the occurence of callose in pollen tubes as they grow down the style may be of systematic interest. As pollen tubes move towards the ovules, the

Fruit, seed and seedling

Fruit type. I define fruit loosely as a post-fertilisation structure developed from either a single carpel, or connate carpels (ovary superior), or connate (or free) carpels (ovary superior) + associated calyx and/or corolla; or connate (or free) carpels + inferior ovary or hypanthium; or ditto + carpels of adjacent flowers; or carpels of separate flowers plus inflorescence axis... A few taxa have more than one fruit type on the plant, a phenomenon known as heterocarpy (Kaul et al. 2000). Products of superior, syncarpous ovaries include a capsule, a dry, dehiscent, syncarpous fruit; berry, which has a ± entirely fleshy pericarp; and a drupe, which is indehiscent and fleshy, but with a more or less woody endocarp enclosing the seeds individually or together ("stones" or "pits"). An achene is a dry, indehiscent, single-seeded fruit with a moderately developed pericarp, and is generally used for the product of a single carpel, but it also used for the fruits of Asteraceae, which have a syncarpous inferior ovary; "cypsela" is a term used to distinguish these fruits from achenes, etc.). A nut is usually larger and with a very strong pericarp or fruit wall; it is generally used when the gynoecium is syncarpous, but the distinction between nut and achene is often not clear in the general literature. A follicle is a single carpel dehiscing down the adaxial side; other terms used for fruits derived from separate carpels include berrylet and drupelet. Fruits that are the products of an inferior ovary are described using the same terms as for fruits developed from syncarpous superior gynoecia; of course, the actual tissues involved in being fleshy may well differ in the two cases. A schizocarp is any fruit that splits into separate units each based on a single carpel and consisting of pericarp plus seed or seeds; these units are dispersed independantly, and/or the seeds may later be released; basically, it is a kind of septicidal/septifragal capsule. A samara is any dry, winged, indehiscent fruit, and a lomentum is a fruit with transverse constrictions along which the fruit later breaks.

The above terms are usually used to describe the products of the gynoecium of a single flower. Another way of describing fruits is to distinguish between a simple fruit, the product of a syncarpous gynoecium, or of a single, free carpel, an aggregate fruit, all the products of separate carpels of a single flower, and a multiple fruit, a unitary structure that is produced from several separate flowers; an anthocarp is a fruiting structure produced from a single flower but which is more or less closely invested by an accrescent calyx or perianth.

Capsules usually dehisce either down the septal or placental radii or down the interseptal or interplacental radii - such fruits are described as being septicidal and loculicidal respectively. In this context, no distinction is made between fruits derived from ovaries with axile placentation and those with parietal placentation. Thus Ericoideae, Cleomaceae and Brassicaceae are all described as having septicidal capsules, even though the first has septae and largely axile placentation, the second, parietal placentation and a unilocular ovary, and the third a false septum and parietal placentation; dehiscence in all occurs in a basically similar position. Dehiscence may also be circumscissile, the fruit opening around the circumference of the ovary or carpel (a fruit with this type of dehiscence is often called a "pyxidium"). In dry, dehiscent fruits with an inferior ovary, the actual dehiscence is often via openings in the pericarp at the top of the fruit, as in many Myrtales, and so dehiscence can reasonably be described using the terms just mentioned. Dehiscence may also occur through the sides of an inferior ovary, as in Campanuloideae, Orchidaceae, etc.; this situation is described when it occurs, but dehiscence is here often not simply septicidal or loculicidal.

I had initially wanted to use terms that reflected whether the fruit was derived from a superior or from an inferior ovary, and if the former, whether the carpels were free or fused. However, some of the terms were cumbersome and others flouted common usage (using separate terms for fruits from apocarpous and from monocarpellary gynoecia - e.g. Leins 2000 - runs into similar problems). Furthermore, a number of fruit types refer to fruit morphologies found only in single families or parts of families ("pepo", "hesperidium"), or imply distinctions when either none really exist or variation is at a lower level than is the focus here (the confusing set of terms "silique", "siliqua", "silicula" can thus be disposed of, which must be something of a relief to all). If more detail is required about how fruits are categorized, extremely complex and detailed sets of terms are available, for which, see Baumann-Bodenheim (1954), Spjut (1994), Roth (1997), etc.; Judd (1985; see also Judd et al. 2002) provides a simplified set of terms.

Whatever problems there may be with the terms used for fruits (see also Clifford & Dettmann 2001) - and many fruits are indeed complex structures - there is much of systematic interest to be gained from their careful study, not to mention relating structure to function in dispersal (van der Pijl 1982). Here I largely restrict myself to the simplified terms suggested above and defined in the glossary, although other information may be added in the characterisations. Thus fruit descriptions in the characterizations must be interpreted carefully, particularly taking into account the position of the ovary, but also the presence of other organs associated with the fruit proper and involved in fruit or seed dispersal. Here, as elsewhere, the terms used to describe fruits are simply needed for descriptive purposes. Thus there may be only very slight differences between the fruit "types", e.g. circumscissile capsule/pyxidium versus achene/utriculus, as Costea et al. (2003) emphasize in a study of the fruits of Amaranthus. Fruit morphology seems rather labile, berries, for instance, having evolved many times, for example, within monocots (e.g. Dahlgren & Clifford 1982; Givnish et al. 2005, 2006b), and capsules may even be derived from berries (Rasmussen et al. 2006).

Although details of fruit anatomy are rarely mentioned, they are clearly important in deciding whether fruits that are ostensibly of the same basic kind really are similar enough to be considered immediately comparable; gross anatomy may be insufficient (Clifford & Dettmann 2001). Thus Bobrov et al. (2005) and others discuss the need to study the developmental origins of the woody layer surrounding the seed - all drupes are not created equal. For a survey of mucilages and gums in fruits, see Grubert (1981).

Whether or not the calyx is deciduous, marcescent, persistent or accrescent (see the fruit type anthocarp) in fruit can be a surprisingly consistent character at higher levels.

Seed size shows very interesting variation when examined across all of angiosperms (Moles et al. 2005).

Details of seed coat anatomy are taken especially from Corner (1976), Krach (1976), Takhtajan (editor: 1985, 1988, 1991, 1992, 2000), Danilova (editor: 1996), Werker (1997), and Bouman and his collaborators (e.g. Bouman & Boesewinkel 1997). As Wunderlich (1967a), Corner (1976), G. Dahlgren (1991) and many others have emphasized, variation in seed coat anatomy is of considerable systematic interest. Fortunately, details of seed coat anatomy can often be observed in hand sections that are stained for lignin, however, which cells are testal and which tegmic in origin can be difficult to see in the mature coat. The testa develops from the outer integument and the tegmen from the inner integument. The prefixes exo- meso- and endo- refer to tissues developing from the outer epidermis, the middle part, and inner epidermis respectively of either the testa or tegment (Corner 1976; see also Schmid 1986). The shape, lignification and inclusions of the cells in these layers is of interest, in particular, much attention has been paid to which part of the seed coat forms the protective or mechanical layer. A malpighian layer of vertically elongated and much thickened cells is quite common, although varying in origin (Werker 1997); it is often a systematically important character. A vascularised (see ovule variants above) and multiplicative testa and/or tegmen (see integument thickness), the latter with the number of cell layers increasing after fertilization, are mentioned only when they occur. In drupe-, achene- or nut-type fruits in particular the seed coat is often undistinguished at maturity and systematically uninformative, the protective function of the seed coat being carried out by the fruit wall. Seed coat anatomy of saprophytes (Bouman et al. 2002) and parasites also may be of little use in detecting relationships since the seeds of such plants are often very reduced, light, and dispersed by wind. Distinctively thickened transfer cells may be found in various places in the seed coat (see e.g. Boraginaceae) or embryo. For a survey of mucilages and gums in seeds, see Grubert (1981). Some monoocts have distinctive substances associated with the testa, in particular, the black phytomelan and brownish phlobaphene (for the former, see Graven et al. 1998; Rasmussen et al. 2006).

Many asterids have only a single integument, and the seed coat of such taxa is simply called a testa in the characterisations. The amount and patterning of the thickening of the walls of the exotestal cells of this single integument often varies greatly within a family, as in Gentianaceae (e.g. Guérin 1904) and Ericaceae. The thickening on the outer periclinal wall is usually less than that on the other walls. However, much of the variation in thickening, cell shape, etc., of this single-layered seed coat is of interest at levels lower than those on which I focus here.

Endosperm development is accompanied by cell wall formation (cellular endosperm), or nuclear division may only later be followed by the laying down of walls (nuclear). In helobial endosperm, the endosperm forms two compartments in one or both of which the protoplasm is initially not divided by wall formation (Wunderlich 1959 for a summary). Krishnamurthy and Indra (1985) summarise discussion over the occurence of helobial endosperm; note that some authors restrict its occurence to monocots. This classification hardly describes the subtleties of endosperm development (Floyd et al. 1999; Floyd & Friedman 2000, 2001). Thus the first division may be variously asymmetric.; one or both of the chambers that result from the first division may be nuclear; etc.

Endosperm presence/absence refers to the situation in the ripe seed, however, it can be a difficult character to deal with. Endosperm is very nearly always found in the very young embryo, but whether or not it persists is of interest here. There is of course a continuum between scanty and copious endosperm, and even the absence of endosperm is by no means as distinct a state as it might appear. Other endosperm variants, such as ploidy levels other than triploid are mentioned as they occur.

Ruminate endosperm is sporadically distributed, from the relatively huge seeds of Myristicaceae to the small seeds of some Scrophulariaceae (it is rare in monocots). Rumination develops in various ways, and characterizes a number of taxa at the lower levels of this survey (Bayer & Appel 1996 for a summary).

Especially in asterids, endosperm haustoria, generally multicellular semi-invasive projections from the micropylar and/or chalazal ends of the seed, are common (e.g. Crété 1951; Mikesell 1990). They are mentioned only where they occur, as are other endosperm variants such as the highly asymmetric early endosperm development of Acanthaceae.

Endosperm food reserves are usually oily and/or proteinaceous, and this is not mentioned; starch, hemicellulose, or amyloid and other polysaccharides (Kooiman 1960; Czaja 1978; Buckeridge et al. 2000) also occur, and the occurrence of such reserves alone is recorded. Starchy seeds are often associated with parasitic and aquatic habits; starchy and/or thick-walled and hemicellulosic endosperm is especially common in monocots (Huber 1969). In monocots with starchy endosperm, as in several Poales and Commelinales, the outer part of the endosperm, the aleurone layer, commonly has other than starchy reserves; this is not mentioned. Starch grains in the young endosperm and in the vegetative plant are commonly of the "Hülle" type, that is, circular when seen from above, and homogenous (Czaja 1978); only variants of this type as found in the seed are mentioned in the descriptions, but this is a difficult character to deal with.

In a few taxa endosperm-like tissue is in fact perisperm, derived from nucellar or similar tissue of the parental sporophyte. Rudall (2000) emphasized the distinction between perisperm and chalazosperm, the latter being derived from the chalaza, not from subdermal nucellar tissue. Even perisperm s. str. originates in various ways, and can develop from the epidermis alone of the nucellus (cf. Acorales). It usually contains starch as a reserve, only rarely being oily; the nature of the reserve, where known, is mentioned in all cases.

Embryo development. Although study of the patterns of cell division in young embryos and of the roles of the cells in the production of the different tissues of the mature embryo has resulted in embryos being placed in a number of "Types" and "Variations" (Johansen 1950; see e.g. Johri et al. 1992, pp. 62-80 for a good summary), there is more than one way to look at these patterns and roles. Johansen (1950, p. 94) observed of the six Types he recognized, "merely to state a prophecy, it is believed that they will survive the test of time". Also, little attention is usually paid to the sometimes quite extensive infraspecific variation in such features. I have not utilised variation in embryo development here. The embryonic suspensor varies considerably in size. De Fulvio (1979) suggests that this feature (along with endosperm type) characterises groups of families (e.g. in the asterid I group), in some taxa, however, it does not develop at all (Wardlaw 1955, for some details; Yeung & Minke 1993, for the physiological significance of the suspensor where the cells can reach 8000 C).

Embryo length varies considerably, but I have not treated this feature critically enough; Martin (1946) is still a very useful source of basic information, while Baskin and Baskin (1998) add much information on various kinds of dormancy of the seed, linking it with embryo size, etc. In both, speculations on the evolution of embryo size, type, and dormancy, should be treated with caution; Forbis et al. (2002) have recently evaluated embryo size in the context of phylogeny. The length of the embryo is emphasized here; it is not absolute, but is the length of embryo relative to the length of the seed. Long embryos are at least half the length of the mature seed and include cases where the embryo is curved and, when stretched out, is much longer than the seed, medium embryos are about half the length of the seed, short embryos are between one half and about one tenth the length of the seed, while minute embryos are still smaller. The states of this character are arbitrarily delimited. Mangrove taxa, and to a certain extent aquatics in general (e.g. Nymphaeaceae, Cetarophyllaceae, Nelumbonaceae), have large and well-developed embryos, perhaps to ensure rapid establishment of the seedling after it germinates. Of course, that embryos are long does not mean that endosperm is necessarily absent, also, if seeds are minute (the "Micro" type of Martin 1946, see also Baskin & Baskin 1998) their contained embryos necessarily will be minute in absolute size, although by the criterion of relative length they might well be long... The embryos of such very small seeds are often undifferentiated (see below, Embryo morphology), while "minute" embryos as defined above often require a period after dispersal for maturation to be completed (morphological dormancy); even if they have cotyledons, etc., they will have to grow to fill up the inside of the seed before germination begins (see also Forbis et al. 2002).

Embryo position. The embryo is more or less central in the seed, the radicle being adjacent to the micropyle, if present, and any endosperm surrounds the embryo. However, in a few families such as Poaceae, Polygonaceae, and also Caryophyllaceae and their relatives (this latter group makes up the old Curvembryonae) the embryo is lateral, lying against the testa on one side and adjacent to endosperm (or perisperm) on the other. Sometimes the position of the embryo relative to the fruit as a whole is of interest, thus in Lamiaceae the radicle points towards the base of the fruit and in Boraginaceae towards the apex - base and apex being relative to the pedicel.

Embryo morphology, that is, the morphology of the embryo in ripe seeds, varies considerably. In the characterizations I mention mainly cotyledon number, whether (and how) the cotyledons are folded, and whether the embryo is straight, curved, and flattened. In embryos derived from curved ovules the cotyledons are accumbent when they are in the plane formed by the long axis of the seed plus chalaza or simply by the curvature of the embryo, incumbent when they are at right angles to this plane; they can (of course!) also be intermediate. In monocots the shape of the embryo varies at about the family level. The relative proportions of cotyledon:hypocotyl + radicle and the thickness of the cotyledon can be of systematic interest; in some Clusiaceae and Lecythidaceae in particular the embryo - sometimes quite massive - is largely hypocotylar. In mycotrophic and/or parasitic plants with minute seeds the embryo is more or less undifferentiated, even the cotyledons not being visible. Larger seeds can also have undifferentiated embryos, thus at most weakly differentiated embryos characterise the entire Cyperaceae - Ericocaulaceae - Poaceae clade, although the embryos of Poaceae themselves represent an extreme in embryo differentiation with the first leaves of the seedling already developed in the embryo, and this is true of Fabaceae-Faboideae, etc.

Data on embryo color are taken largely from Yakolev and Zhukova (1980), but see also Janzen (1982) and Wright et al. (2000). The sampling is poor, but some taxa commonly have green embryos, others have white embryos; the latter is the common condition. Seubert (1993) suggests that in monocots green embryos are found only in Alismatales. Only Viscum and its relatives (Santalaceae) have a green endosperm.

Cotyledon number is usually an unambiguous character, although in taxa with two cotyledons like Cyclamen, some Ranunculaceae, etc., that produce subterranean storage organs the cotyledons may be more or less connate. Burger (1998) and others have suggested that the single cotyledon of the monocots might be equivalent to the third leaf of the Nymphaealean seedling, and that the young embryo of both Nelumbo and Nymphaeales is initially monocot-like. Unfortunately for this idea, Nelumbonaceae, Nymphaeaceae and monocots are not now thought to be immediately related... For the literature on monocotly in broad-leaved angiosperms, and suggestions that Syneilesis (Asteraceae) seems to lack cotyledons entirely, see Teppner (2001). Polycotyledony is particularly noteworthy in Pinus (as in pine "nuts"), but also occurs rarely elsewhere, e.g. Idiospermum (Calycanthaceae).

Cotyledon position. The cotyledon(s) are usually lateral structures; only in monocots, with the exception of Poaceae, do they appear to terminate the embryonic axis.

Seed germination and seedling morphology provide a wealth of systematically interesting characters, but I have attempted to integrate these systematically into monocot descriptions only. In hypogeal seedlings the epicotyl is elongated, while in epigeal seedlings it is the hypocotyl that is elongated; cryptocotylar seedlings have non-photosynthetic cotyledons that remain enclosed by the seed coat and/or pericarp during germination, while in phanerocotylar seedlings the cotyledons are exposed and photosynthetic (e.g. de Vogel 1980, who develops a complex typology of germination patterns, and also gives references to other such endeavours). The terms hypogeal and epigeal as used in descriptions of monocot seedlings (e.g. in Kubitzki ed. 1998a, c; Linder & Caddick 2001) largely correspond to the terms crypto- and phanerocotylar used to describe the seedlings of broad-leaved angiosperms. However, I have used neither pair of descriptors above in the monocot characterisations, since in a monocot seedling the cotyledon is commonly exposed and photosynthetic in its lower part, while the apex remains enclosed by the testa/fruit wall and is absorbtive, and any hypocotyl present is at most short by comparison to that of a broad-leaved angiosperm with epigeal germination.

Arber (1925), Boyd (1932) and Huber (1969) in particular provide much data for seedlings in monocots, and Tillich (e.g. 2000, 2003) suggests that, with a few exceptions, there is little variation in seedling morphology within monocot families. In many monocots, and some aquatic broad-leaved angiosperms, the radicle is poorly developed or aborts; abortion of the radicle also occurs in some Cactaceae (Rodríguez-Rodríguez et al. 2003). There is frequent infrageneric variation in broad-leaved angiosperms in particular as to whether or not the hypocotyl is developed, and also whether or not the cotyledons are photosynthetic (e.g. Burtt 1974). The shape of the exposed cotyledons can be of systematic interest, as in e.g. Bignoniaceae and Burseraceae, and also, in monocots, whether the cotyledon is unifacial or bifacial. Cotyedonary vasculature can vary considerably within a family, and/or differ when compared with that of adult leaves (Stone 1970).

Phyllotactic patterns of very young plants are interesting. In monocots the first leaf is generally at 180° to the cotyledon, whereas in broad-leaved angiosperms it is at right angles to the plane of the cotyledons. Henslow (1893 and references) discussed some of the abrupt shifts in phyllotaxis in monocot seedlings, while in broad-leaved angiosperms the first leaves may be opposite, even when other leaves of the plant are distichous or spiral (e.g. some species of Phaseolus), and of course the leaves of seedlings can have a morphology very different to that of the adult leaves.

NUCLEUS

Chromosome numbers are taken largely from secondary sources, including the various indices of chromosome numbers. All counts have been silently converted to haploid numbers (n); the occurrence of polyploidy within families, etc., is rarely mentioned. Although endopolyploidy is rife in flowering plants, with 24,576n estimated for the endosperm haustoria cells of Arum maculatum (Bennett 2004), this is also not mentioned except in certain situations, e.g. tapetal cells (see above). I also rarely mentioned the literature in which base numbers (x) for families or orders are suggested, but Raven (1975) is a convenient entry into this field. There is increasing evidence for genome duplication, perhaps through polyploidization, at various times during the evolution of seed plants (see below). Furthermore, there is evidence from isozyme duplication of polyploidization of clades like Magnoliaceae, Aesculus, Salix/Populus, etc. (Soltis & Soltis 1990). The notably small stomatal size of some fossils when compared with extant members of clades also suggests polyploidization within these clades, some extinct members perhaps having chromosome numbers half of any of those known (Lauraceae, Magnoliaceae and Platanaceae: Masterson 1994). Working out the evolution of chromsome numbers thus becomes difficult (and note that the examples just mentioned are all from woody groups that are unlikely to have had herbaceous ancestors; given our current understanding of the problem, polyploidization is less likely in such groups...); Meyers and Levin (2006) provide a general overview of the consequences of polyploidization.

Karyotype. Detailed features of the karyotype have been used to characterise a number of monocot (in particular) groupings at around the family level in Asparagales and Liliales (Tamura 1995 summarises much information). Although I may have made less of such characters than I might (but some details are being added to the pages), they are not notably conservative. For instance, although a bimodal karyotype characterises an Agavaceae that includes genera that used to be in Hostaceae and Hyacinthaceae-Chlorogaloideae, etc., a broadly circumscribed Agavaceae for some people's tastes (but to be included in Asparagaceae s.l.), it is also found elsewhere in Asparagales. Most chromosomes have telomeres, and it was thought that telomere construction was rather invariant, the sequence that was repeated being very conservative. However, it has recently been found that there is quite a bit of variation that is systematically interesting at various level from individual genera on up, e.g. within Asparagales and within Solanaceae (e.g. Adams et al. 2001; Sýkorová et al. 2003a, b, 2006).

Genome size, whether measured as the 1C or C-value (unreplicated gametic genome) or Cx value (unreplicated basic genome), shows considerable variability. Leitch et al. (2005) summarise what is known about this for all seed plants (see Leitch et al. 2001 for gymnosperms) and and Bennett and Leitch (2005) extend this to all land plants. The variation they detail is largely consistent with phylogeny; seed plants have small genomes (less than 1.4 pg) compared to those of most gymnosperms (intermediate, 3.5-14 pg). Others are assembling lists of genome sizes in angiosperms (e.g. Hanson et al. 2005; Zonneveld et al. 2005). There is often no correlation between C measures and ploidy level (e.g. Weiss-Schneewiess et al. 2005). For variation in the length of the mitotic cycle, see Evans and Rees (1971); the length seems to be associated with the degree of coiling of the chromosomes.

Nuclear inclusions such as protein bodies in nuclei of various shapes and sizes are occasional (e.g. Thaler 1966; Speta 1977, 1979 for information); they are mentioned only when they occur. Other nuclear inclusions in asterids are described by Bigazzi (e.g. 1995). Note that some literature describing such inclusions dates back to the late nineteenth century (see Speta 1979).

There is increasing evidence that genome duplications have been quite common in the history of land plants (e.g. Vision et al. 2000; Bowers et al. 2003; Blanc & Wolfe 2004; Schlueter et al. 2004; Maere et al. 2005; de Bodt et al. 2005; Chapman et al. 2006); these may be at the level of generic groupings, as well as deep within Poaceae or Brassicaceae, the common ancestor of the asterid and rosid clades, most angiosperms, and even of all seed plants. Indeed, potential duplications seem remarkably common (Cui et al. 2006), although exactly where many are to be put on the tree remains unclear because of limited sampling. Some of the gene duplications, duplications of individual genes, that are often mentioned in the literature may be linked to these genome duplications (note that rates of gene duplication alone are appreciable - Lynch & Conery 2000). Both can be invaluable sources of phylogenetic information and clues for understanding diversification and the evolution of morphology; the functions of the duplicated genes may initially remain similar, but subsequently there is often subfunctionalisation, neofunctionalisation, or loss of one of the genes (for a nice example, see de Martino et al. 2006). Much work has been done on the duplication of the PHY genes (genes for phytochrome variants); Mathews et al. (2003) follow the early evolution of the PHYA and PHYC genes (see also Mathews et al. 1995; Mathews & Sharrock 1996, etc.; for general background see Karniol et al. 2005 and especially Mathews 2006b [the big picture of phytochrome evolution in an ecological context]), while Schmidt and Schneider-Poetsch (2002) discuss the complex pattern of duplication of the PHY genes in the gymnosperms. The duplication of the RPB2 gene (the gene coding for the second largest subunit of RNA polymerase II) provides important phylogenetic information (Oxelman et al. 2004). For the duplication and distribution of the LEAFY gene, see Frohlich and Parker (2000) and Frohlich (2002).

Variation in floral developmental genes (also linked to genome duplication, of course) is of great evolutionary and phylogenetic importance. The ABC model, developed in the context of studies on the core eudicots Arabidopsis and Antirrhinum in particular, suggests a fairly close linkage between particular classes of genes and their expressions in particular organs of the flower, but it is clear the in the Anita grade, magnoliids, etc., expression is not so localised (e.g. Li et al. 2005; Kim et al. 2005). A genes are found only in flowering plants. In more basal groups of flowering plants, the pattern of AP3 expression is not always consistent and its level changes during development, but in core eudicots the expression becomes spatially restricted within the perianth. The euAP3 gene in angiosperms consists of a duplicated copy of the paleoAP3 gene with an eight BP insertion causing a frameshift mutation (Vandenbussche et al. 2003). In Asarum and monocots AP3 expression is localised on the edges of the tepals. A-class APETALA 1 genes also are of considerable interest (Litt & Irish 2003). For the distribution of SEPALLATA genes in angiosperms, see Zahn et al. (2005) and Malcomber and Kellogg (2005). APETALA 3 and similar genes and PISTILLATA are paralogous B-class genes (e.g. Stellari et al. 2004). For the duplication of B-function MADS-box genes, see Kim et al. (2004: see also Kramer & Irish 1999, 2000; Kramer et al. 1998; Theißen et al. 2000, 2002). C/D genes in gymnosperms are represented by AG-like genes that are expressed in micro- and megasporophylls, as well as ovules. There was a duplication event in the ancestor of angiosperms that gave rise to a D orthologue, expressed largely in the ovules alone, and a C orthologue, only sometimes expressed in ovules, but also in carpels and stamens. Both patterns of gene duplications within the C lineage and patterns of relationships within individual gene lineages (for which, see Kramer et al. 2004) seem to correlate with particular phylogenetic hypotheses. Patterns of duplication in such developmental genes can be taxonomically quite fine-grained (Kramer et al. 2003, 2004), thus Howarth and Donoghue (2005) have begun to untangle the patterns of gene duplication of CYC genes in Dipsacales and link it with morphological change there.

PLASTID AND MITOCHONDRION

For surveys of chloroplast inheritance, whether via the pollen grain or the egg cell, see Corriveau and Coleman (1986) and Mogensen (1996); there is much variation in gymnosperms.

Some major genome rearrangements are mentioned. Thus the chloroplast inverted repeat characterises most land plants and a few of their immediate relatives (e.g. Turmel et al. 1999). Downie and Palmer (1992b) summarise early literature on the use of chloroplast DNA rearrangements in reconstructing phylogeny, and there is much information on later work in Raubeson and Jansen (2005). Data on the loss of the intron in the chloroplast rpl2 gene is taken from Downie et al. (1991), that in the rpoC1 intron from Downie et al. (1996), and in rpl16 gene from Campagna and Downie (1998). Millen et al. (2001) summarize information about the loss of the infA gene from the chloroplast and Joly et al. (2001) that about the loss a coxII.i.3 intron from the mitochondrion. For the loss of the ndh genes in the chloroplast of Pinus thunbergii, see Wakasugi et al. (1994). Adams et al. (2001, 2002a, b) follow the loss of a number of genes from the flowering plant mitochondrial genome, probably because of their transfer to the nucleus; the rps13 gene however, seems to have been replaced in many Rosids by a chloroplast gene that is now in the nucleus. In a number of cases, fairly high-level groupings appear to be marked by gene migration, the loss of the rps2 and rps11 genes perhaps being particularly significant systematically (see also Bergthorsson et al. 2003), although in other cases individual genera (or perhaps groups of genera) such as Phlox, Allium, etc., show more extensive migration of genes. Graham et al. (2000) discuss microstructural changes in noncoding DNA; the distributions of these changes support clades such as all angiosperms, the eudicots, monocots minus Acorus, and Austrobaileyales. Y.=L. Qiu and Palmer (2004) have suggested that how mitochondrial introns are spliced (cis or trans) may be of systematic significance (see Cameron et al. 2003).

There seems also to have been significant horizontal transfer of genes, although the extent of this is unclear. Mitochondrial genes in particular seem to have been transferred easily, and in some cases this is linked with the close association of host and parasite, genes moving from the latter to the former (Davis & Wurdack 2004; Mower et al. 2004). Other cases involving transfer of genes between angiosperms and gymnosperms and even bryophytes and angiosperms are even less understood (Won & Renner 2003; Bergthorsson et al. 2004).

Finally, there is support for many of the relationships suggested here from analyses of the variation in molecular sequence data; these are cited separately in the appropriate places on the order pages. Not all sequences are equal, with mitochondrial ITS sequences perhaps being particularly difficult to use (Álvarez & Wendel 2003).

OTHER

Host-plant preferences, particularly of butterfly larvae (e.g. Forbes 1958; Ehrlich & Raven 1964; Ackery 1988, 1991; Farrell et al. 1992; Powell et al. 1999; Ward et al. 2003) can provide information indirectly bearing on relationships, although interpretation may be difficult and there is no recent synthesis of the huge and scattered literature, made more difficult because temperate and tropical studies (see Janz & Nylin 1998; Novotny & Basset 2005; Lewinsohn et al. 2005); Fielder (1997) compared host plant utilisation by temperate and tropical butterflies. Larvae of individual species may also show interesting host-plant preference patterns. There is a valuable database of caterpillar host plants that I have consulted - see Hosts - and associated with this are conventional books that detail food preferences of the caterpillars of parts of the area covered in the database (Robinson et al. 2001, 2002); see also papers in Scriber et al. (1995) and especially Berenbaum (1995) for swallowtails (Papilionidae) and their foodplants.

The patterns of fungal parasite/host associations are also often of some systematic interest, and Savile (1979b) provides a comprehensive and appropriately cautious summary of these associations. Since both fungi and insects that eat or parasitise plants are often affected by the chemistry of their hosts, parallelisms in chemistry may be reflected by finding related fungi, or butterfly or moth groups, (for example) on plants with similar chemistries. Thus members of the unrelated pairs magnoliids and Rutaceae (both have similar alkaloids), Putranjivaceae and Brassicales (glucosinolates), and Daphniphyllaceae and asterids (iridoids) have caterpillars of related butterflies eating them. Along the same lines, that similar caterpillars are found e.g. on both Onagraceae and Vitaceae (Forbes 1956) may reflect the fact that both groups have raphides.

The ability of grafts of different species, genera, and even families to take also probably reflects something of the underlying physiology of the plants concerned. However, the literature dealing with graft compatabilities is very scattered, and grafts between very distantly related plants seem to be possible (e.g. see Horne 1914; Hartmann 1951).